搜档网
当前位置:搜档网 › PAPER Disparity Mapping Technique and Fast Rendering Technique for Image Morphing

PAPER Disparity Mapping Technique and Fast Rendering Technique for Image Morphing

PAPER Disparity Mapping Technique and Fast Rendering Technique for Image Morphing
PAPER Disparity Mapping Technique and Fast Rendering Technique for Image Morphing

PAPER

Disparity Mapping Technique and Fast Rendering Technique for Image Morphing

Toshiyuki MORITSU?,Nonmember and Makoto KATO?,Member

SUMMARY We have developed a new disparity mapping technique for image morphing which prevents synthesized images from blurring and a fast rendering technique which realizes in-teractive morphing animation.In the image morphing rendering process,all pixels are moved according to their disparity maps and then distorted images are mixed with each other.Calcula-tion costs of this process tend to be high because pixel per pixel moving and mixing are included.And if the accuracy of the disparity maps is low,synthesized images become blurred.This paper describes new two techniques for overcoming these prob-lems.One is a disparity mapping technique by which the edges in each input image are accurately mapped to each other.This technique reduces blurring in synthesized images.The other is a data transformation technique by which the morphing render-ing process is replaced with texture mapping,orthographic cam-era,α-brending and z-bu?ering.This transformation enables the morphing rendering process to be accelerated by3D accelerators, thus enabling interactive morphing animations to be achieved on ordinary PCs.

key words:image morphing,image-based rendering,view in-terpolation,interactive animation

1.Introduction

Image morphing is one of the most powerful visual e?ect techniques.It creates a series of intermediate images from several digital input images.

Recently,view morphing has garnered a great deal of attention and a lot of research has been done on this technique[4],[6].View morphing is a view interpola-tion technique which uses morphing.View morphing doesn’t require any3D models,which is an important feature because there are many objects for which the construction of3D models is di?cult,such as animals, plants,clothes and architectures.

1.1Quality of Synthesized Images

In the morphing technique,the quality of the synthe-sized images mainly depends on the accuracy of the disparity maps.The disparity maps are maps which indicate pixel per pixel mapping between input images. If their accuracy is low,the feature points of the input images don’t move to the same position in the synthe-sized images,and thus these images become blurred. The more input images are used,the more serious this Manuscript received March19,1999.

Manuscript revised August4,1999.

?The authors are with Systems Development Labora-tory,Hitachi Ltd.,Kawasaki-shi,215–0013Japan.problem becomes.

In most existing morphing systems,disparity maps are calculated via a semiautomatic process.In these systems,the operator speci?es corresponding features in each input image by using straight line segments[5], and the disparity maps the other pixels are calcu-lated automatically by using these line segments.In this technique,however,there are many features which don’t appears on the line segments.and thus the dis-parity maps for them must be surmised.This may cause mismatching of these features and make the syn-thesized image blurred.

In this paper,we assume that the main reason synthesized images become blurred is mismatching of edges.Following this assumption,we show a new tech-nique which enables us to create maps of edges in input images accurately and e?ectively.In our technique,we divide input images into several regions along the edges and map the borders of the regions of each input image accurately.

1.2Rendering Speed of Image Morphing

There are two types of image morphing rendering pro-cess:the warping process and the cross-dissolve pro-cess[5].In the warping process,all input images are distorted according to their mixing rates and their dis-parity maps.Mixing rates are speci?ed at rendering time and they control the similarity of synthesized im-ages to each of the input images.In the cross-dissolve process,all the warped images are mixed according to their mixing rates.

The pixels in all input images are moved in the warping process,and the pixels in the warped images are mixed with each other in the cross-dissolve pro-cess.If the size of the input images becomes large, the process cost becomes high and it becomes di?cult to realize interactive rendering with an ordinary MPU. In ideally,speci?c morphing hardware is required.But hardware only for morphing is too limited in usage,and thus this isn’t a realistic solution.

We think one of the best solutions is to render mor-phing data with existing3D accelerators because they enable fast rendering and their rendering processes re-semble morphing rendering processes.In this paper, we show a technique which transforms morphing data into data which can be rendered on ordinary3D accel-

erators.By our transformation,the warping process in morphing rendering process can be done through tex-ture mapping and the cross-dissolve in morphing ren-dering process can be done throughα-blending.

We have used the technique shown in this paper and light position interpolation technique[2]to develop a system for exhibiting Noh masks,which are used in Japanese traditional drama.An indication of the suc-cess of the system is that a video of the Noh mask, made using it,was accepted for inclusion in the SIG-GRAPH’98Electronic Theater Program?and was in-vited to be included in NICOGRAPH’98Film Show. And preliminary version of this work was presented at Paci?c Graphics’98[1].

The remainder of this paper is structured as fol-lows:In Sect.2,we show our technique of disparity mapping.In Sect.3,we show transformation of mor-phing data for rendering on3D accelerators.In Sect.4, we show our future work.We conclude in Sect.5.

2.Disparity Mapping Technique

In this section,we describe a disparity mapping tech-nique for reducing blurring in synthesized images.

2.1Basic Strategies

We make two assumptions regarding the cause of blur-ring in synthesized images:(1)If the edges of each in-put image aren’t mapped to each other accurately,the synthesized images will tend to be blurred.(2)Mis-matches of the smooth parts in each input image aren’t critical.In line with these assumptions,we adopt the following strategies:

1.Accurately map the edges of each input image each

other

2.Create the maps semiautomatically

3.Create the maps of the smooth parts automatically 2.2Steps for Disparity Mapping

In line with these strategies,we design the following steps and develop tools for them.In the remainder of this paper,we use the input images shown in Fig.1as an example.

1.Regional division along with edges in input images

Much research has been done on regional division along the edges of images[9],[10],and a number of retouch software products have regional division capabilities.By using one of these products,we divide input images into several regions.

Example:

In our example,we divide the input images into 8regions:hair,face,neck,clothing,left eye,right eye,nose,and mouth(Fig.2).The regions are de-termined as follows:

?An area with an edge on the border is a region.

?An area that has a disparity map that is com-

pletely di?erent from that of the surrounding

areas is a region.

2.Specify line segments which cross the borders of

the regions

The operator speci?es line segments which cross the borders of the regions in one of the input im-ages.The operator also speci?es corresponding line segments in one of the other input images.

Next,we determine pixels on which the line seg-ments and the borders of the regions cross.We call these pixels cross points,and we assume that the cross points speci?ed by corresponding lines are mapped to each other.

Example:

Figure3shows the steps for calculating the

dis-Fig.14input images which we use as

examples.

Fig.28Regions which we divide input images into.

?You can see the video in the SIGGRAPH Video Review ISSUE:125No.7“Noh Mask-Application of Image-based Rendering.”

MORITSU and KATO:DISPARITY MAPPING TECHNIQUE AND FAST RENDERING TECHNIQUE FOR IMAGE MORPHING

277

(a)

(b)

(c)

(d)

(e)

(f)

Fig.3(a)Two pairs of corresponding line segments are spec-i?ed in input images.(b)Disparity maps of pixels which are on the borders are calculated.(c)Several pairs of corresponding line segments are added to get more accuracy.(d)All disparity maps of pixel which are on the borders of this region are generated.(e)Other features are speci?ed with line segments.(f)Interpolated image.

parity maps of the face region.In Fig.3(a),two pairs of corresponding line segments are speci?ed in each input image.The white points on them are cross points.

3.Calculate disparity maps of all the other pixels which are on the borders

Disparity maps of all the other pixels which are on the borders of the regions are calculated with these cross points as follows.First,we select two cross points which are on the same border and are next to each other.We call these points P ointA and P ointB .Next,we ?nd corresponding points of P ointA and P ointB in one of the other in-put images.We call these points P ointA and P ointB .Next,we ?nd the line segment which runs along the border and whose ends are P ointA and P ointB .We call this line segment LineAB .In the same way,we ?nd LineA B .Finally,we as-sume the intermediate points of lineAB and those of LineA B ,which divide LineAB and LineA B in the same rate,are mapped to each other.Example:

The white points on the border in Fig.3(b)are some sampled points which have disparity maps and are used by the operator to ?nd incorrect maps on the border.The white points which are on the

same order of distance from the cross points are mapped to each other.These maps are generated by the cross points which speci?ed in Fig.3(a).In Fig.3(b),the white points on the forehead are mapped properly.At the chin,however,we can ?nd incorrect maps.In these cases,more accurate maps can be generated by specifying additional pairs of corresponding line segments (Fig.3(c)).In this facial region case,we also specify line segments which cross the other borders;i.e.,those with left eye,right eye,nose and mouth (Fig.3(d)).4.Specify disparity maps of other features

There are some features which are hard to divide into regions,like scratches or stains.We specify these features by using points and line segments.Ideally,these features should also be speci?ed with edge tracing lines,but we have omitted this func-tion this time.Example:

In Fig.3(e),we specify facial moles with short line segments.

5.Calculate disparity maps of all the other pixels Disparity maps of all the other pixels are calcu-lated by using the pixels for which disparity maps were already generated in the above steps.This algorithm is shown in Fig.4.With this algorithm,the pixels in one of the input images (image )are mapped to those in one of the other input im-ages (image ).The program list is written in C++like grammars.In the program list,Words in typewriter face indicate reserved words and function calls.Words in roman face indicate ab-stracted functions.Words in italic face indicate variables.PositionOf (pixel )is a function which calculates the coordinates of pixel in the image.MappedPixelOf (pixel )is a function which ?nds mapped pixels of pixel in image .R is a con-stant variable by which the calculation range of unmapped pixels is de?ned.

One of the pixels (pixel )in image which doesn’t have a disparity map is selected at step 1.The variables totalDistortion ,totalInvDistance and distance are de?ned at steps 2,3and 4.One of the pixels (pixel m )in image which has a disparity map is selected at step 5.The distance between pixel m and pixel is calculated at step 6.If distance is longer than the value which speci?ed with constant variable R ,we don’t use this pixel m to determine the disparity map of pixel at step 7.Distortion of pixel m is added to totalDistortion at step 8.The inverse of distance ,which controls pixel m ’s in?uence,is added to totalInvDistance at step 9.At step 11,if there aren’t any pixels which have a disparity map near pixel ,we determine the dispar-ity map of pixel later.We determine the disparity map of pixel as a weighted average at step 12.

278

IEICE TRANS.INF.&SYST.,VOL.E83–D,NO.2FEBRUARY2000

1.while(pixel in image which doesn’t have a disparity map exists){

2.var totalDistortion=(0,0);

3.var totalInvDistance=0;

4.var distance;

5.for(each pixel m in image which has a disparity map){

6.distance=||PositionOf(pixel m)-PositionOf (pixel)||;

7.if(distance≥R)continue;

8.totalDistortion+=PositionOf(MappedPixelOf

(pixel m))-PositionOf(pixel m);

9.totalInvDistance+=1/distance;

10.}

11.if(totalInvDistance==0)continue;

12.PositionOf(MappedPixelOf(pixel)))=PositionOf (pixel)+totalDistortion/totalInvDistance;

13.}

Fig.4Algorithm which calculates disparity maps of all the other pixels.

2.3Result:Synthesized Image

We generate synthesized images in almost the same way as when an ordinary morphing algorithm is used.First, warped image for each input image is generated by transforming the input images.The position of each pixel in the input images is linearly interpolated ac-cording to the disparity maps with all the other input images and to the mixing rates which are de?ned for each input image at rendering time.The synthesized image is then generated by mixing the warped images in the mixing process.

The image in Fig.5(a)is synthesized from4in-put images which is generated by using above rendering algorithm and the image in Fig.5(b)is actual photo-graph which taken at almost same position at which the synthesized image indicates.These input images are mixed at the same rate(25%each).Despite the mixing of the4images,the synthesized image doesn’t become blurred.The image shown in Fig.5(c)is mag-ni?ed portion of the image in Fig.5(a).

And the images of a reindeer?gure in Fig.6are another example of our rendering algorithm.The im-age in Fig.6(b)is synthesized from4input images in Fig.6(a),and the image in Fig.6(c)is an actual pho-tograph.In this example,some di?erence between in synthesized image and in actual photo can be found.At ?rst,the legs of the reindeer get blurred and the face of reindeer is rounded.These di?erences are caused by occlusion.And this occlusion problem is going to be pointed at Sect.4.

3.Transformation of Morphing Data for Ren-

dering on3D Accelerators

In this section,we show a technique which transforms morphing data into data which can be rendered on or-dinary3D

accelerators.Fig.5(a)Synthesized image created from4input images at the same mixing rates.(b)Actual photo.(c)Magni?ed portion of the image in

(a).

Fig.6(a)Input images of a reindeer.(b)Synthesized image created from4input images in(a).(c)Actual photo.

3.1Basic Strategies

Most current3D accelerators have texture mapping and α-blending capacities.Texture mapping is a kind of image distortion process andα-blending is a kind of image mixing process.Since the warping process re-sembles texture mapping and the cross-dissolve process resemblesα-blending,we developed a data transforma-tion technique which the warping process can be done through texture mapping and the cross-dissolve process can be done throughα-blending.

3.2Morphing Data

Before showing the data transformation process in de-tail,we show the original morphing data.In this paper,

MORITSU and KATO:DISPARITY MAPPING TECHNIQUE AND FAST RENDERING TECHNIQUE FOR IMAGE MORPHING

279

we use the structured morphing technique[3].which technique is suitable for view morphing.In this tech-nique,input images are divided into several layers and each layer is distorted individually.This achieves natu-ral deformation both with near objects and far objects and also enables easy treatment of occlusions.In our example,we divide the input image into the following three layers.

1.Hair

2.Face(including eyes,nose,and mouth)

3.Body and neck

This layer division is done by unifying the previ-ously segmented regions.

3.3Warping by Using3D Accelerators

In this subsection,we explain the data transformation in which the warping process is replaced by texture mapping.In ordinary3D accelerators,texture images are distorted by using polygons,after which they are mapped to synthesized images.We use a triangle as the primitive?gure for distortion.The data transfor-mation consists of two steps,one in which each layer is divided into triangular patches,the other in which distortion in these patches is determined by using a disparity map.

Distortion resulting from the use of triangular patches may cause unexpected distortions and blurs. Unexpected distortions mainly occur in parts where im-age distortion changes drastically,and blurs mainly oc-cur in parts where edges are included.We use?ner patches for these parts to reduce the distortional e?ect. The division is done as follows:

1.Divide each layer into equilateral triangles

We divide each layer into equilateral triangles of identical size.The size is speci?ed by the operator.

Example:

Figure7shows the layer-division process.Fig-ure7(a)shows one of the original layers.In Fig.7(b),the layer has been divided into equilat-eral triangles.

2.Subdivide patches containing an edge or edges or

in which distortion changes drastically

We subdivide patches into4small equilateral tri-angles if they contain an edge or edges or in which the distortion changes drastically.For example, consider patches2and6in Fig.8(a).These patches are subdivided as shown in Fig.8(b).

Whether a patch contains an edge or edges or not depends on whether or not it contains the board-ers of the regions that were determined by the pre-ceding regional-division step.Whether or not the distortion changes drastically depends on whether or not the patch contains pixels whose di?eren-tial value about disparity map exceeds the speci-?ed threshold

value.Fig.7(a)Original image of a layer.(b)The layer is divided into equilateral triangles.(c)Patches in which distortion change drastically or in which edges are exist are subdivided.(d)Patches are deformed with their vertex’s movements.(e)Texture image is mapped on distorted patches.

Fig.8(a)Original8patches.(b)We subdivide patches2and 6into4?ner patches.(c)Patch1is subdivided into4?ner patches because it has two subdivided patches as neighbors.(d) Patches3,5and7is subdivided into2?ner patches because they have one subdivided patch as a neighbor.

3.Subdivide a patch if2or3neighboring patches are

subdivided into?ner patches

If patches are divided into?ner patches,cracks may appears due to the distortion in the patches.

These cracks may appear when a vertex of a di-vided patch boarders an undivided patch.To avoid this problem,we subdivide patches that are next to subdivided patches.We subdivide a patch into 4small equilateral triangles if2or3neighboring patches are already subdivided into?ner patches.

We repeat this step until no patch that is not sub-divided has2or3subdivided neighboring patches.

For example,patch1in Fig.8(a)is subdivided into 4?ner patches because it has2subdivided neigh-boring patches(Fig.8(c)).

4.Subdivide a patch if a neighboring patch has been

subdivided into?ner patches

280

IEICE TRANS.INF.&SYST.,VOL.E83–D,NO.2FEBRUARY2000

We subdivide a patch into2small triangles if a neighboring patch has been subdivided into?ner patches.For example,patches3,5,and7in Fig.8(a)are subdivided into2?ner patches be-cause there is a subdivided neighboring patch (Fig.8(d)).

Example:

In Fig.7(c),the patches in which the distortion changes drastically or in which there are edges and their neighboring patches are subdivided by using the above algorithm.

Next,we show the steps by which distortions in patches are calculated.We identify the distortions in patches by using the movements of their vertices.These movements are linearly interpolated by using the mix-ing rates of their input images and the disparity maps of their vertices.The disparity maps are determined as follows:

1.The temporal disparity maps of the vertices are

determined by using one patch

When we doing this,two cases have to be consid-ered.The?rst case is that the vertices are included in the layer(vertices shown in Fig.7(c)).In this case,we determine that the disparity maps are the same as those of the pixels at the same position.

The second case is that the vertices aren’t included in the layer(vertices not shown in Fig.7(c),which exist just outside the layer).In this case,we de-termine that the disparity maps are approximately the same as those of the nearest pixels in the patch.

2.The average disparity maps of the vertices between

the patches that include the vertices

If the vertices are included in multiple patches,we calculate the average disparity maps between these patches and regard them as the?nal values.If the vertices are included in only one patch,we regard the maps calculated in step1as the?nal values. Example:

The disparity maps for each of the patches shown in Fig.7(c)are calculated using by the two steps.The dis-torted patches are shown in Fig.7(d),and the patches with texture mapping are shown in Fig.7(e).

3.4Mixing by Using3D Accelerator

In this subsection,we explain the data transformation in which the cross-dissolve process is replaced byα-blending.

We assume a simple3D space containing an ortho-graphic camera and plates(Fig.9).

1.Orthographic camera

The orthographic camera projects plates.

2.Plates

The plates are transparent.The warped image of each layer of each input image is pasted on a

plate Fig.9Simple3D space for executing cross-dissolve process on 3D accelerators.

at the appropriate transparent rate.The plates are stacked vertically in front of the camera in layer or-der.The order of the plates within each layer is not important.The transparent rate of the plates,ex-cept for the backmost plates of each layer,is deter-mined by the mixing rate of the input images from which the warped images are created.The trans-parent rate of the backmost plates of each layer al-ways become zero(meaning the warped images are pasted opaquely).The relationship of each layer to the layer in front and in back of it is main-tained through z-bu?ering.The warped images of the input images can be mixed with each other by α-blending.

3.5Dynamic Input Image Selection

As the number of input images is increased,rendering speed becomes slow.In our example,we have already used4input images.If we want to change the facial expression,more input images are required.We there-fore added a capacity for dynamic input image selec-tion.This technique limits the number of input images which is used in one rendering cycle.This limitation is speci?ed by the operator,and when the rendering process is executed,speci?ed number of input images with high mixing rate are automatically selected and rendered in accordance with the mixing rate of each of the input images.

3.6Hardware Requirement

Our rendering technique requires the following capaci-ties in3D Accelerators.

?Texture mapping

?Orthographic camera

?Paste image to plate with mask

?α-blending

?z-bu?ering

MORITSU and KATO:DISPARITY MAPPING TECHNIQUE AND FAST RENDERING TECHNIQUE FOR IMAGE MORPHING

281

Fig.10Synthesized images.Each image is generated in 82msec with an ordinary PC.

To the best of our knowledge,most3D APIs have these capacities.Some accelerators,however,don’t support all of them,and we also found that acceler-ators with a simpli?edα-blending capacity don’t work particularly well.

3.7Result:Synthesized Image and Performance

The images shown in Fig.10are generated from4input images by using the above rendering technique.Ren-dering time is82msec per image;thus we can accom-plish interactive morphing animation at a rate of about 12frame/sec.The measurement environment is as fol-lows.MPU:Pentium?II266MHz,Memory:96MByte, OS:WindowsNT??4.0,3D API:OpenGL???,3D accel-erator:Graphic board powered by Permedia????2,Im-age size:410×400pixels,Number of dynamic input image selections:2images(i.e.,2input images used in one rendering cycle).Under nearly identical condi-tion(without Permedia2),we executed ordinary mor-phing rendering(Pixel per pixel moving and mixing with MPU).We synthesized images from two input im-ages(because our new rendering technique uses two input images for one rendering cycle),and the time re-quired was920msec per image.Thus,by using the rendering algorithm described in this paper,rendering speed becomes about11times faster.

4.Future Work

Future work includes improving the system so that it can map pixels in one input image to occluded pixels in the other input image(in case of occlusion).In our system,all the pixels in a region of an input image are always mapped to the pixels in the corresponding region in the other input image.If one region is partly or fully hidden by the other region so that the corresponding pixels are occluded,we can’t map these pixels correctly. For example,in Fig.6(b),the legs of the reindeer are ?Registered trademark of Intel Corp.

??Registered trademark of Microsoft Corp.

???Registered trademark of Silicon Graphics Corp.

????Registered trademark of3Dlabs Corp.blurred and the face of the reindeer is rounded because the legs and the side of face are hidden in some of the input images.To mitigate these e?ects,we need to use more input images and to use mapping methods that specify the position of occluded pixels.

The system also needs to be improved so that im-ages whose contents are hard to divide can be mapped correctly.A method is needed to calculate the dispar-ity maps by using corresponding line segments along the edge.

5.Conclusion

This paper has described new two techniques relevant to image morphing.One is a disparity mapping tech-nique by which edges in each input image are accurately and e?ectively mapped to each other.This technique reduce blurring in synthesized images.The other is a fast rendering technique for image morphing.In this technique,the warping process is done through tex-ture mapping and the cross-dissolve process is done throughα-blending.By this transformation,the mor-phing rendering process can be accelerated with the use of ordinary3D accelerators;we were able to achieve 12frame/sec(Image size is410×400)interactive mor-phing animation on an ordinary PC with an inexpensive 3D accelerator.

Currently,we are developing exhibition systems for valuable works of art,and there are many works for which construction of3D models is di?cult.We are using the morphing technique to change the view points of these works.For high end use,synthesized im-ages with high quality are required.And when we use morphing as view interpolation,the number of input images is tend to be increase.If we do two dimensional movement,at least3input images are required.Under these conditions,an accurate disparity mapping tech-nique is strongly to be required to create high quality (less blurred)synthesized images.

In the?eld of interactive CG animation,most all applications use geometry-based technique.Image-based rendering has garnered much attention recently, but currently,the main?elds for this technique involve non-interactive use,such as in movies and TVcom-mercials.As this paper indicates,however,the use of an interactive image-based rendering technique has good possibilities for expanding the application?elds for image-based rendering.

References

[1]T.Moritsu and M.Kato,“Disparity mapping technique and

fast rendering technique for view morphing,”Proc.Paci?c Graphics’98,pp.216–217,Singapore,Oct.1998.

[2]K.Hisaki and M.Kato,“Changing lighting positions

in photographs without3D geometric modeling,”Proc.

VSMM’98,Gifu,Japan,Nov.1998.

[3]M.Kato and H.Noyama,“Interactive visual simulation in

IEICE TRANS.INF.&SYST.,VOL.E83–D,NO.2FEBRUARY2000 282

a quasi-three-dimensional world based on structuralization

of images,”ICAT/VRST’95,pp.101–110,1995.

[4]S.E.Chen and L.Williams,“View interpolation for image

synthesis,”Proc.SIGGRAPH’93,pp.279–288,Aug.1993.

[5]T.Beier and S.Neely,“Feature-based image metamorpho-

sis,”Computer Graphics Proceedings,pp.35–42,July1992.

[6]S.M.Seitz and C.R.Dyey,“View morphing,”Proc.SIG-

GRAPH’96,pp.21–20,1996.

[7]M.Okutomi and T.Kande,“A multiple baseline stereo,”

IEEE Trans.Pattern Anal.&Mach.Intell.,vol.15,no.4,

April1993.

[8] D.Scharstein and R.Szeliski,“Stereo matching with

non-linear di?usion,”https://www.sodocs.net/doc/d42705415.html,puter Vision and Pattern

Recognition,pp.343–350,June1996.

[9] E.N.Mortensen and W.A.Barrett,“Intelligent scissors for

image composition,”Proc.SIGGRAPH’95,pp.191–198,

1995.

[10]M.Kass,A.Witkin,and D.Terzopouls,“Snakes:Active

contour models,”1st.International Conference on Com-

puter Vision,pp.259–268,1987.

[11]H.Tanaka and F.Kishino,“Adaptive mesh generation for

surface reconstruction:Parallel hierarchical triangulation

without discontinuities,”https://www.sodocs.net/doc/d42705415.html,puter Vision and Pat-

tern Recognition,pp.88–94,1993.

Toshiyuki Moritsu was born in To-

kyo,Japan,in1969.He received B.E.and

M.E.degrees in computer science from

Keio University in1993and1995.He is

presently with Systems Development Lab-

oratory,Hitachi Ltd.His research inter-

ests includes image processing,computer

graphics.He is also involved in devel-

oping techniques for the securities settle-

ment systems.

Makoto Kato received his B.Liberal

Arts degree in1979,and M.S.in physics

in1981,respectively from the University

of Tokyo.He joined Systems Develop-

ment Laboratory,Hitachi,Ltd.in1983.

He is interested in image processing,com-

puter graphics.He is also involved in

developing techniques for electronic com-

merce systems.

猎头服务合同书

编号:_______________ 本资料为word版本,可以直接编辑和打印,感谢您的下载 猎头服务合同书 甲方:___________________ 乙方:___________________ 日期:___________________

委托方____________________以下简称甲方代理方_________人才服务部以下简称乙方甲方因业务发展需要,委托乙方搜索所需人才。 经双方友好协商,达成以下合作条款第一条 主旨本合同目的在于确定乙方代表甲方搜索候选人时双方的权利及义务。 第二条 搜索服务的基本方式1甲方需提供与所需搜索职位有关的详细资料给乙方,其中主要包括--工作环境,如公司背景、现状规模、发展状况、目前和将来的投资等;--职务名称、职位职责,个人发展前景、直属上司情况及汇报路线;--所须具备的条件包括年龄、性别、教育程度、英文程度、专业经验以及个人品性;--工作中所需要的独立性、工作氛围,流动性出差需要和上班地点等;--福利待遇包括基本工资、奖金和津贴、工作时数、生活条件以及合同保险。 2乙方工作程序--搜索可能达到该标准的候选人,对这些候选人进行初次面试筛选并确证其资料的真实性;--将候选人资料提交甲方参考,包括对其经验的评价及对其性格,能力和潜质的看法;--为双方安排面谈时间、地点,并参与其中一些条款的协商,协助聘用合同的最终签署;--协助甲方督促被录用候选人与原工作单位按正常的程序办好离职手续。 3乙方在合同签定之日起天内完成对甲方所需人才的寻访、测评及背景调查,将完整、真实的候选人资料提交甲方,并安排候选人面试。 4甲方应在收到乙方提供的候选人资料后两日内,做出是否与己有人才资料重复的判断,否则视为乙方推荐;甲方应在收到乙方提供的候选人才资料后一周内,做出是否需要复试的判断,甲方应在候选取人面试后两周内将面试意见及审核结果反馈给乙方,在复试后四周内做出是否录用的判断。 第三条 保证期1推荐的候选人被甲方录用,并最终到甲方工作,视为猎头服务成功。 2雇员在保证期三个月内无论因任何原因离职,甲方应五天内向乙方提出书面报告,乙方将免费为甲方推荐另一合适候选人到职;乙方收到甲方书面报告后二个月内没有推荐合格人选到职,乙方应退还该项服务费的_______。

猎头协议(标准范本)

猎头协议 In accordance with the relevant provisions and clear responsibilities and obligations of both parties, the following terms are reached on the principle of voluntariness, equality and mutual benefit. 甲方:__________________ 乙方:__________________ 签订日期:__________________ 本协议书下载后可随意修改

协议编号:YH-FS-825220 猎头协议 说明:本服务协议书根据有关规定,及明确双方责任与义务,同时对当事人进行法律约束,本着 自愿及平等互利的原则达成以下条款。文档格式为docx可任意编辑使用时请仔细阅读。 甲方:_________ 乙方:_________ 1.公司指定 根据本合同协议和条款,乙方委托并指定甲方为乙方提供本合同规定的各项咨询服务。 2.双方关系 甲方和乙方均为独立法人。本合同的签订在甲方和乙方之间并不产生任何雇用、代理、合资或业务伙伴关系。甲方绝非乙方的法律代表,无论是任何目的,甲方均不能以书面或其它明示或隐含方式,以乙方的名义享受权利或承担义务。 3.服务 甲方将根据乙方提供的职位要求向乙方推荐合适的人选。具体职位描述、要求及人数由乙方提供,详细的书面材料作为本协议的附件一,具有同等法律效力;

甲方应确保每一个推荐给乙方的候选人首先经过甲方的面试筛选。若乙方要求,甲方应对候选人简历中包含的信息或候选人提供的其它信息的准确和真实性及在面试过程中发现的问题进行核实。 凡经甲方推荐的人才,从推荐之日起12个月内,如果乙方未通过甲方暗地或通过第三方录用该人才,无论长期或短期,均视为甲方推荐成功,乙方按照本合同相关条例支付咨询费用。如果乙方把甲方推荐的人才中转推荐给第三方,甲方亦按照本合同相关条例向乙方要求支付咨询费用; 4.机密 甲方承诺: 1)对获知的乙方及其分支机构的商业和技术信息保守秘密; 2)对获知的乙方或分支机构的雇员或候选人相关信息保守秘密。 3)甲方推荐的候选人及其身份和职业相关信息只可提供乙方使用,不得向乙方以外的第三方泄露候选人的信息。 乙方接受本合同后,也将承诺所有由甲方推荐的候选人

猎头合同模板

合同编号: 高级人才寻访服务合同 甲方: 地址: 电话: 邮箱: 乙方: 地址: 电话: 传真: 甲、乙双方在平等自愿、协商一致的基础上,就甲方委托乙方寻访和推荐优质人才等事宜,达成如下条款,以共同遵守: 一、服务内容 1.1甲方委托乙方按照甲方提供的职位描述要求寻访人才,接受乙方的推荐,并安排面试。 1.2乙方通过各种渠道搜寻符合甲方职位要求的合适候选人,并推荐给甲方,以满足甲方人 力资源方面的需求。 1.3当乙方所推荐的候选人通过甲方面试,并在甲方所指定的岗位上报到后,该服务主体结 束。但是乙方有义务提供合同所承诺的人员更换服务义务。 二、服务期限 本合同自双方签字起生效,合同有效期为年,自年月日至年月日。 三、甲方的权利和义务 3.1 该职位委托合同甲方委托乙方为其代理招聘服务,。 3.2 甲方须尽可能详细明确的填写《委托猎头登记表》,以及提供委托职位的职位描述。甲方应根据项目进度及乙方服务规程要求,给予乙方项目顾问尽可能详尽的信息支持与配合,以便乙方寻访适合甲方的候选人; 3.3甲方应在乙方推荐人员到职后三个工作日内书面通知乙方。如在聘用通知书约定的上班之日,乙方推荐的人员未能到甲方上班的,甲方应将该情况发生后当日内书面通知乙方;3.4甲方应在与乙方推荐人员签订聘用通知书后五个工作日内将所聘人员职位、到职日期、薪酬、试用期等情况以书面形式通知乙方; 3.5甲方应对乙方提供的人员资料保密,并严格控制内部人员知情范围,不得向外泄露人才资料,或向任何第三方提供人才资料(包括个人和法人)。如有上述情况,一经核实,甲方需无条件给予乙方相应的经济赔偿;

3.6甲方对乙方的任何投诉和建议均可通过发送邮件到邮箱,甲方的投诉和建议将反馈给乙方相关部门。 四、乙方的权利和义务 4.1 甲方填写委托登记表、签订委托招聘服务合同后,乙方开始正式提供合同约定服务; 4.2 乙方应在签订合同后的10个工作日内,向甲方提供书面或电子邮件形式的,与委托职位相适合的有关候选人资料报告; 4.3 乙方自签订本合同起,不得将甲方内的工作人员、商业信息,推荐或泄露给其他可能获得利益的第三方;在本协议到期或终止后,以上保密责任依然存在。 4.4 乙方不得将其推荐给甲方并被成功聘用的人选再次推荐给其它任何客户; 4.5 乙方承诺提供人员更换服务如下:如乙方推荐的候选人在正式报到日起3个月内离职,甲方在被推荐人选离开之日起10个工作天日内以书面形式将该情况通知乙方,乙方应就此职位按甲方要求无偿提供另外人选。新人选入职后,乙方将重新计算保证期并继续承担人员更换服务义务。如果被录用者的工资福利、岗位职位等未能按甲方与候选人事先约定之标准实施,或者遇甲方因办公场所搬迁、重大人事调整所导致的非人选自身原因而导致被录取者离职的,由甲方承担对该被录取者的相关责任,乙方不承担人员更换服务义务; 4.6 任何乙方向甲方所推荐的人员,自被推荐之日起十二个月内,均被视为“乙方推荐人员”,在此期间内,如乙方证实甲方或其子公司与被推荐人发生聘用与被聘用关系,视为乙方已成功推荐,甲方须按本合同支付全额费用; 4.7 自“乙方推荐人员”被推荐予甲方之日起十二个月内,甲方不得就聘用事宜向任何第三方(不包括甲方各分支机构)推荐。 4.8 在合同期内,乙方不得向甲方在职员工推荐任何职位机会。 五、服务费用及付款方式 5.1 成功聘用乙方推荐的候选人后,甲方须向乙方按每人每一职位支付服务费用标准为被雇佣者第一年税前年薪总收入的。被雇佣者年薪总收入包括基本工资、奖金、提成及其他各类现金支付的福利和补贴; 5.2 本合同所称的成功聘用有下述两种情况: 1)甲方与乙方推荐的候选人签订了《劳动合同》或者该候选人员已实际在甲方工作 (含试用期); 2)包括但不限于甲方与乙方推荐的候选人以顾问或者兼职的形式产生合作关系。 5.3 服务费支付方式如下:乙方推荐的候选人被甲方成功聘用后乙方应向甲方发出付款通知书,甲方应在收到付款通知核实无误后的十个工作日内支付服务费用的。候选人通过三个月试用期之后,乙方应向甲方发出付款通知书,甲方应在收到付款通知核实无误后的十个工作日内支付服务费用的30%。如金额有误,甲方应在收到付款通知后两个工作日内通知乙方;乙方未收到甲方书面的异议通知的,视为甲方确认该付款通知书。 5.4 乙方在提供本合同约定的服务时实际发生的所有费用,如广告费,差旅费等须经甲方事先书面同意,方可由乙方向甲方收取。如甲方延期付款超过天,乙方有权要求甲方及时付清服务费,每天按服务费的加收滞纳金。

猎头协议正式版

The cooperation clause formulated through joint consultation regulates the behavior of the parties to the contract, has legal effect and is protected by the state. 猎头协议正式版

猎头协议正式版 下载提示:此协议资料适用于经过共同协商而制定的合作条款,对应条款规范合同当事人的行为,并具有法律效力,受到国家的保护。如果有一方违反合同,或者其他人非法干预合同的履行,则要承担法律责任。文档可以直接使用,也可根据实际需要修订后使用。 甲方:_____ 乙方:_____ 1.公司指定 根据本合同协议和条款,乙方委托并指定甲方为乙方提供本合同规定的各项咨询服务。 2.双方关系 甲方和乙方均为独立法人。本合同的签订在甲方和乙方之间并不产生任何雇用、代理、合资或业务伙伴关系。甲方绝非乙方的法律代表,无论是任何目的,甲方均不能以书面或其它明示或隐含方式,

以乙方的名义享受权利或承担义务。 3.服务 甲方将根据乙方提供的职位要求向乙方推荐合适的人选。具体职位描述、要求及人数由乙方提供,详细的书面材料作为本协议的附件一,具有同等法律效力; 甲方应确保每一个推荐给乙方的候选人首先经过甲方的面试筛选。若乙方要求,甲方应对候选人简历中包含的信息或候选人提供的其它信息的准确和真实性及在面试过程中发现的问题进行核实。 凡经甲方推荐的人才,从推荐之日起12个月内,如果乙方未通过甲方暗地或通过第三方录用该人才,无论长期或短期,均视为甲方推荐成功,乙方按照本合同相

猎头服务合作协议书

猎头服务合作协议书 甲方(委托方): 地址: 乙方(受托方): 地址: 甲、乙双方本着互利互惠、平等自愿、有偿服务的原则,就甲方委托乙方代为搜索、甄选、推荐各类种人才等事宜,经充分协商,达成如下协议,以资共同遵守: 一.合同有效期:从______ 年—月—日至__________ 年—月—日。 二.甲方的权利与义务 1. 甲方必须向乙方提供本企业营业执照复印件、公司简介、委托招聘的相关信息。 2. 甲方应按乙方要求提供所需的职位需求信息,并保证寻猎职位信息的真实性,所提要求及待遇 标准,必须按标额执行。如在委托期内,寻猎需求有任何更改应及时通知乙方。 3. 甲方负责对乙方提供的候选人资料进行筛选,对符合初选条件的人,乙方负责协调候选人到甲 方进行面试。 4?如甲方发现乙方提供的人选资料与其他招聘渠道所提供的资料重合时,且甲方从乙 方处获得候选人姓名、联系信息之前,或要求乙方安排候选人面试之前,应在收到乙方人选资 料后5个工作日内提出申明,并出具相关的资料证明,否则即视为认可。 5. 甲方按照公司 的相关面试流程对乙方推荐的候选人进行面试,在面试流程结束后,5 个工作日内告知乙方面试结果。 6. 甲方按时向乙方支付服务费。 7. 在本协议有效期内或本协议终止后十二个月内,甲方聘用乙方推荐的人才的,甲方应于聘用前通知乙方,并按 照本协议第3条规定向乙方支付服务费。甲方聘用乙方推荐的人才但未按前述规定履行通知义务的,甲方除应向乙方一次性支付全额服务费外,应向乙方支付服务费的两倍作为违约金,且甲方聘用的 乙方推荐人才的年薪视为30万元,甲方不得对此提出任何异议。若乙方发现其推荐的人才的年薪高于30 三.乙方的权利与义务 1. 乙方在双方签订委托招聘服务合同后,明确招聘岗位后,寻访和物色工作全面正

猎头合同协议书律师审核版

猎头合同协议书律师审 核版 文稿归稿存档编号:[KKUY-KKIO69-OTM243-OLUI129-G00I-FDQS58-

人才服务协议书 合同编号:XXXXXX 甲方(委托方): 地址: 联系人及电话:传真: 乙方(受托方):XXXX 地址:XXXX 联系人及电话:传真: 甲乙双方经协商一致,就甲方委托乙方在全国范围内代为招聘、甄选、推 荐人才事宜,经充分协商,达成如下协议,以资共同恪守。 一、服务内容 1.甲方委托乙方寻访的职位要求、数量、寻访期、签约年薪(用以约定 寻访委托金)、寻访委托金、寻访服务费等事项,可在由甲方不定时 发出的推荐要求中载明。该推荐要求应以书面形式发出,并作为本协 议的附件。 2.乙方根据甲方提出的人才需求,按甲方要求在约定的时间内为甲方寻 访候选人。经乙方评估认为符合甲方岗位要求后,向甲方推荐,经甲 方面试合格、候选人上岗后,甲方按本合同的规定向乙方支付约定的 服务费用。 二、服务流程 1.在双方签订本协议后,乙方正式启动为甲方提供的猎头服务。 2.在向乙方开放猎头职位的2个工作日内,甲方应当用电子邮件形式向 乙方所需猎头职务的详细职位说明(含工作职责、任职资格、薪金范 围、基本要求、特别需求等),以便乙方搜寻、筛选和向候选人说明

情况等,甲方保证寻访职位信息的真实性;如在委托期内,寻访需求有任何更改均应及时通知乙方。 3.乙方在收到甲方提供的书面职位说明后,应在2个工作日内以电子邮 件形式予以确认,并与甲方进行沟通确认;经双方对职位需求确认一致后,乙方开始进行人才访寻及推荐。 4.自对职位说明书确认后之1个工作日内,乙方指派专人或小组负责为 甲方提供人才寻访推荐服务,以保证按时完成甲方的招聘需求,并在10个工作日内提供第一批候选人至少2-3名,甲方应在收到推荐报告资料3个工作日内向乙方提供对所推荐候选人的反馈意见(如安排面试、暂缓、拒绝),并将同意面试的候选人名单以电子邮件形式提供给乙方,对于需要安排面试的候选人,甲方须在15个工作日内予以面试;乙方根据甲方要求及名单约见面试,甲方在面试后的10个工作日内将面试结果以电子邮件形式通知乙方,如甲方对参加面试的候选人不满意,应以电子邮件形式反馈明确的原因给乙方,乙方应重新搜寻合适的人选,具体工作程序同上。 5.如果乙方向甲方提供的简历已被甲方从其它渠道获取,甲方应当在3 个工作日内以电子邮件形式通知乙方,如乙方提出要求,甲方应提供相应的书面证明。如在上述期限内甲方未提出任何异议的,则该候选人将被视为乙方提供的资源。 6.如甲方对乙方提供的候选人有聘用意向的,应甲方要求及候选人同意 的情况下,乙方应提供候选人背景核查,核查应从公开途径且不侵害候选人隐私。

猎头公司服务合同

猎头服务合同 委托单位(以下简称甲方): 代理单位(以下简称乙方): 甲方因业务发展需要,委托乙方猎聘所需人才,经双方友好协商,达成如下合约: 第一条招聘内容 第二条甲方的权利义务 1、甲方必须向乙方提供真实详细的企业背景资料和招聘的《职位描述报告》。 2、甲方对乙方推荐的候选人进行面试后,必须在3个工作日内给出是否录用该候选人的意见。 3、甲方不得利用候选人获取候选人违反其与原工作单位签订的保密协议规定的保密义务,因甲方利用候选人 获取原单位保密信息而引起的任何后果由甲方独自承担,与乙方无关。 4、甲方作为雇主应当对自己雇员的最终选择、录用、持续工作监管承担全部责任,对于推荐候选人所引起(或 相关)的对甲方造成的损失、损坏或费用等将由甲方自行承担,乙方不承担任何相关责任。 5、合约履行过程中,如甲方变更招聘其他岗位,或取消该岗位的招聘要求,或延缓聘用适用该岗位的候选人 的,乙方有权单方面解除本合约,预付款不予退还。 6、甲方承担候选人面试所产生的所有差旅费用,定向猎取或乙方随同甲方外出洽谈、面试,甲方须为乙方支 付定向所在地猎聘往返差旅费。 7、对于乙方所推荐的候选人的相关资料,甲方不能作为其他用途并负有保密的义务,否则赔偿相应损失,甲 方若将乙方推荐的候选人转给其它第三方并产生聘用的,视为乙方已完成服务工作并支付相关服务费用。 第三条乙方的权利义务: 1、乙方承诺保证不猎取甲方在职人员,并保守甲方公司的机密。 2、乙方按照甲方《职位描述报告》进行人才搜寻、面试,并对甲方公司的招聘信息严格保密。 3、乙方为甲方提供详尽的猎头人才评估推荐报告,为甲方面试提供专业的岗位人选和用人建议。 4、乙方协助甲方与候选人进行薪酬谈判、到岗日期等促进招聘工作,协助甲方与人才签定劳动合同,并进行 后期人才的跟踪服务,如果被录取者的工资福利等未能按甲方与人才事先约定之标准实施,而导致被录取 者离职的,乙方不退还预付款和服务费,并由甲方承担由此而来的一切责任。 5、乙方在服务期内(60个工作日)内未能向甲方推荐候选人或候选人不符合《职务描述报告》要求,甲方有 权终止本合约,乙方退还本岗位全额已收预付款。 6、乙方推荐的候选人,甲方面试后因某种原因不能长期或专职聘用该候选人,而委托该候选人担任兼职或 其它短期服务,同样视为乙方已完成服务工作,甲方须支付相应的服务费用。 7、本协议签订后,甲方鼓动、引诱、劝说乙方推荐的候选人绕开乙方私下签订劳动合同或变更劳动合同的 薪酬标准的,视为乙方已完成服务工作,甲方须支付相应的服务费用。

猎头服务合同模板

猎头服务合同模板文件编号TT-00-PPS-GGB-USP-UYY-0089

企业招聘服务合同 甲方:安捷利(番禺)电子实业有限公司 地址:广东省广州市南沙区环市大道南63号 联系人:温小姐 /罗先生 电话: / 邮箱:传真:乙方:地址:负责人:电话:邮箱:传真: 甲乙双方经友好协商,就乙方为甲方提供招聘服务相关事宜,一致达成如下合同条款,以共同遵守: 一、服务内容 1. 甲方委托乙方按照甲方提供的职位需求,为其进行人力资源引进推荐服务。 2. 本合同中所述委托是指甲乙双方书面(包括但不限于下述形式:合同、协议、信件、传真、电子邮件等)确认甲方需要乙方猎寻的职位以及该职位所需要的关键要求,上述委托可为本合同附件,与本合同具有同等法律效力。

3. 合同期限为:本合同有效期为一年,自签字盖章之日起生效,每年合同到期前经双方书面约定可延续本合同一年。 二、甲方权利义务 1. 甲方须向乙方提供真实的企业相关背景资料。 2. 甲方须配合乙方对推荐的候选人进行面试。 3. 甲方应在与乙方推荐人员签订聘用通知书后三个工作日内,将所聘人员职位、到职日期、年薪、试用期等情况以书面或电子邮件形式通知乙方。 4. 甲方应在乙方推荐人员到职后三个工作日内以书面或电子邮件形式通知乙方。 三、乙方权利义务 1. 乙方严格按照甲方提供的岗位说明书对人才进行搜寻,每个岗位提供的候选对象不得少于三人。 2. 乙方对初选合格人员进行面试并核准相关证明资料和工作经历。 3. 乙方负责与甲方沟通确认甲方面试的时间、地点,并进行相关人才的联络服务工作。

4. 乙方协助甲方进行人才聘用合同的最终签定。 5. 乙方承诺提供候选人更换服务如下: 1) 甲方需支付完乙方服务费。 2) 如乙方推荐的候选人在正式上岗日起保证期三个月内离职,甲方在被推荐人离开之日起三个工作日内以书面或电子邮件形式通知乙方,乙方为甲方免费提供一次相同岗位的人才推荐至成功聘用服务。 3) 乙方在补充推荐新候选人期间,甲方自行找到合适候选人,或者乙方没寻访到合适候选人替代,乙方承诺将相当于该职位服务费的50%用于冲抵合同期内其它成功招聘职位的相应金额服务费,如果合同期满双方未再有成功招聘的职位而造成无法冲抵,则乙方承诺将该职位服务费的50%退还给甲方。 4) 如果被录用者的工资福利、岗位职位等未能按甲方与候选人事先约定之标准实施,或者遇甲方因办公场所搬迁、重大人事调整所导致的非人选自身原因而导致被录取者离职的,由甲方承担对该被录取者离职的相关责任,乙方不承担人员更换服务的义务。 6 . 本合同期内,乙方不得向甲方在职员工推荐任何职位机会。

猎头服务协议书模板

编号:猎头服务协议书模板 甲方 乙方 签订日期年月日 (本协议书为Word格式,下载后可根据您的需要调整内容及格式,欢迎下载。)

甲方:____________服务有限公司 ____________人才服务中心 乙方:________________________ 甲乙双方就甲方向乙方提供人才猎头服务,达成如下协议: 一、根据中华人民共和国的法律和法规,甲方确认本方有完全的从事本协议所述服务的资格;乙方确认本方是依法成立的组织,符合法律规定的面向社会公开招聘的各项条件和手续,并且向甲方提交的各种书面文件及陈述都是真实有效的。甲乙双方就各自的确认承担相应的法律责任。 二、甲方服务的收费标准是,每成功提供一人,收取________元人民币作为本方的服务佣金。佣金分两次支付,双方签定协议之日,甲方支付________元/人为预付佣金;推荐人选上岗三个月或转正后一个星期内付清余款。 三、甲方收到乙方预付的预付服务佣金后,应按照本协议附件 1的要求为乙方提供人才,并保证所提供人才情况的真实性,同时协助乙方安排有关面试考核的事宜。面试考核应本着方便应聘者及乙方的原则进行。 四、乙方应按照本协议附件1规定的条件考核验收甲方提供的人才。并应于考核之日起________天内将是否聘用(含试用,下同)的意见及理由书面通知甲方。否则,视为同意聘用。 五、甲方应于收到预付服务佣金之日起两个星期至一个月内完

成委托事项,经双方协商同意,上述期限可以改变,但应另以书面文件规定之。 六、乙方同意聘用(接纳为乙方正式员工)甲方提供的人才,甲方服务即告完成。乙方若于聘用后三个月内辞退甲方提供的人才,应于辞退之前通知甲方,并有权要求甲方重新提供服务,甲方应免费为乙方再推荐人选,并得到乙方聘用为止。 七、非因不可抗力,或乙方认可的其他正当理由,甲方在协议规定的期限内未能向乙方提供候选人才的,应全额返还乙方预付佣金;虽提供候选人才而乙方最终未能聘用的,预付佣金应退还甲方。 八、乙方以超出本协议附件1以外的条件拒绝甲方提供的人才,或在本协议规定的有效期限内自行招聘协议附件1规定的人才,从而无需甲方的服务,甲方有权留置乙方全部预付佣金,折抵甲方服务收入。 九、乙方拒绝甲方提供的人才,从而导致甲方服务失败,但在甲方服务终止后的十二个月内聘用曾被乙方拒绝的人才,应于聘用前通知甲方,并向甲方支付全额服务佣金。否则视为违约,乙方应就此向甲方支付________至________倍的服务佣金作为违约金。 十、本协议在乙方委托寻猎的人数超过一人时,对每一个人才的服务分别有效,彼此无法律上的连带责任关系。 十一、本协议的附件作为协议的组成部分,与本协议有同等法

猎头服务合同模板

编号: HT-20214765 甲 方:______________________________ 乙 方:______________________________ 日 期:_________年________月_______日 猎头服务合同模板 The parties may dissolve the contract upon consensus through consultation.

[标签:titlecontent] 甲方(劳务用工单位): 工商登记号:联系地址:邮政编码:联系电话:传真:联系人:开户行:帐号: 乙方(劳务招聘单位): 工商登记号:联系地址:邮政编码:联系电话:传真:联系人:开户行:帐号: 甲、乙双方就乙方向甲方提供需求职位人才的寻访推荐服务达成如下合同(以下称“本合同”): 前言: 甲乙双方保证各自为依据中华人民共和国的相关法律法规合法成立的组织。乙方应确认有从事该项服务的资格,甲方应确认向乙方提供的各种资料的真实性,双方就各自的确认承担各自的法律责任。 甲、乙双方本着自愿和互惠互利的原则,合作签署“本合同”,并愿意自觉遵守“本合同”的各项条款。 第一条服务内容 1、甲方聘请乙方作为其中、高级人才推荐的供应商。

2、甲方如有招聘需求,应向乙方提供委托招聘职位信息,内容包括需求岗位职责描述、胜任条件、薪资福利条件、需求人数和日期等。 3、自甲方正式委托职位寻访之日起,乙方正式启动寻访推荐工作。15个工作日内向甲方推荐每个职位2—3名经过甄选基本符合甲方职位需求的候选人,直到甲方录用乙方所推荐的候选人为止。 4、乙方在推荐候选人时,需要向甲方提供人选的详细简历,说明其以往的工作经历、主要职责、成就以及能说明其工作胜任能力的任何其它必要资料。 5、甲方聘用乙方推荐的人员,在个月(保证期),自推荐人开始在甲方工作之日起,如发生聘用终止,乙方应为甲方免费提供一次相同职位的人才服务,乙方应保证在甲方通知乙方的15个工作日内为甲方再次提供候选人。 第二条限制 1、甲方不得提供虚假职位需求信息和做出虚假承诺。 2、合作期间甲方不得直接聘用乙方猎头顾问和工作人员,如有发生甲方应付乙方该职位应付猎头佣金的双倍的服务费用。 第三条保密原则 1、乙方应对甲方提供的任何商业、技术资料及员工信息进行保密。 2、甲方应对乙方提供的人选资料、服务费用保密,不得将乙方提供的人选资料用于任何非自己聘用内部员工的目的和提供给甲方

猎头合作协议范本

猎头合作协议范本 委托人:(以下简称甲方) 地址: 电话: 电子邮件: 受托人:(以下简称乙方) 地址: 电话: 电子邮件: 甲方委托乙方代理招聘人才,就有关事宜订立合同,目的在于确定甲乙双方的权利及义务,共同遵守如下条款: 一、甲方的权利和义务

1、甲方同意委托乙方猎聘________职位,猎聘佣金按该猎聘职位年薪的______%计算,该职位的猎聘佣金最低不得低于______元。该职位推荐周期为______个月。 2、甲方必须提供营业执照复印件、单位简介等所需文件协助乙方工作。 3、甲方如在乙方安排的初猎面试时不满意乙方报荐人才,可要求乙方重新推荐人才面试。 4、为保证乙方的工作顺利进行,甲方不能单独约见应聘者,面试由乙方安排见面,如有特殊情况约见应聘者,也必须通知乙方。 5、待甲方通过面试选定人才到岗后,人才到岗______个工作日内付猎头服务费用的______%,人才通过试用期后______个工作日内付其余______%的猎头服务费用。如甲方延期付款,每天按余款的百分之______作为滞纳金。 6、甲方在收到乙方发送的候选人资料当日,须以邮件形式反馈给乙方是否通过其他方式获得此候选人简历,如果当日没有回复给乙方信息,并私自约见、录用候选人,甲方除应支付乙方猎头服务费外,同时应承担猎头服务费的______%违约金,并在乙方得知此事后______个工作日内支付。若逾期支付,滞纳金每天为猎头服务费的 ______%。

7、在合作期内,甲方一经决定录用乙方推荐的候选人,应提供录用通知书复印件予乙方。否则,乙方有权利对该候选人进行其他推荐。 8、通过电话沟通,符合甲方职位要求的外地来甲方面试候选人如果要求才旅费的,由甲方承担。 二、乙方的权利和义务 1、乙方从双方签订协议起,寻访和物色工作全面正式开始,并且在每______个工作日内向甲方提交拟举荐的候选人的综合推荐简历(每一岗位的候选人为______名),直至甲方找到合适人选或本推荐周期结束。根据甲方反馈意见,乙方负责安排候选人与甲方面谈具体事宜,同时全程协助及引导性服务。 2、若乙方在受托后向甲方举荐的人选不符合甲方要求,乙方应继续举荐,直到甲方满意或甲方提出终止寻访通知为止;甲方提出终止寻访应以书面的形式通知乙方(如果甲方通过其他方式找到合适的候选人,甲方也应及时以书面形式通知乙方提出终止寻访)。 3、在乙方推荐的人选被甲方正式录用后,乙方承诺______年内不到甲方单位内猎取任何在职人员给其他客户。 4、乙方推荐给甲方的人才享有个______月保用期,保用期内候选人离开(包括但不限于辞职、辞退),甲方有权选择由乙方继续在

人才猎头合同范本

人才猎头服务协议 甲方:xxxxxxxxxxxxxx 乙方:xxxxxxxxxxxxxxx 甲方因业务发展需要,特委托乙方以猎头服务方式招聘职位人选,甲乙双方本着“平等合作,互惠互利”的原则,经友好协商达成如下协议: 一、甲方权利与义务 1.甲方应向乙方提供详细、真实的公司背景资料,包括但不限于公司背景、经营状况、发展战略、企业文化等。 2.甲方须负责提供所需招聘岗位的详细资料,真实填写《企业职务需求表》(附件1),并对薪酬、福利、休假等与应聘者利益相关一切信息之真实性负责。 3.收到乙方提交的候选人报告资料后,甲方应在三个工作日内通知乙方是否要求候选人面试。如果面试,应提前告知时间和地点。面试结束后,甲方应在三个工作日内告知乙方面试结果。对不合适的候选人,应给出说明。如果不面试,应给出具体的不面试原因。 4.甲方应在面试后积极与乙方沟通、协调,三天内做出上一轮面试的决议(包括安排下一轮复试或录用)。特殊情况外,每个候选人面试次数最多为四次。 5.甲方对候选人信息、候选人报告、候选人面试过程等一切可能有损于候选人正当利益或本合同正当履行的信息,负有保密义务。 二、乙方权利与义务 1.乙方对甲方提供的企业、职位等信息负有保密义务。除为完成本合同所必须外,其余未经甲方授权,不得公开。 2.乙方需在寻访服务期(自年月日至年月日)内利用各种渠道为甲方寻访、招聘所需人才。乙方在了解候选人的情况,分析其背景资料,并在此基础上进行筛选

后,向甲方提交合适候选人的个人资料。如果在寻访期内,甲方未能对乙方提供的候选人给出明确答复,乙方有权终止合作。 3.若出现以下任何一种情况,均视为乙方已完成猎头服务全部工作,乙方有权要求甲方按本合同所列的“全职服务费用”,按约定支付周期与方式支付费用。如逾期支付,乙方将保留诉诸法律解决的权利: (1)甲方面试乙方所推荐的候选人后录用其作为甲方员工以全职合作; (2)甲方面试乙方所推荐的候选人后因某种原因不能提供所聘用岗位,但录取该候选人为其他岗位职员。 4.若出现以下任何一种情况,均视为乙方已完成猎头服务全部工作,乙方有权要求甲方按本合同所列的“全职服务费用”,在五个工作日内支付全额猎头服务费。如逾期支付,乙方将保留诉诸法律解决的权利: (1)甲方将乙方所推荐的候选人转荐给其他雇主,并且候选人被他雇主其聘用的; (2)甲方面试乙方所推荐的候选人后表示拒绝聘用,却在面试后一年内聘用乙方曾推荐的人选。 5.若出现以下任何一种情况,均视为乙方已完成猎头服务全部工作,乙方有权要求甲方按本合同所列的“兼职服务费用”,在五个工作日内支付全额猎头服务费。如逾期支付,乙方将保留诉诸法律解决的权利: (1)甲方面试乙方所推荐的候选人后因某种原因不能聘用该候选人,而聘用该候选人做兼职或其他短期服务; (2)甲方面试乙方所推荐的候选人后并没有聘用,而采用咨询、承包等方式与候选人保持合作关系。 三、服务费标准及支付方式 1.全职服务费用标准: □人民币万元(大写:)(岗位年薪标准)×%(年薪的百分比)。 □约定固定金额服务费用人民币万元。 说明:支付方式按双方协商约定,二选一。 若甲方提出的岗位薪酬是某一区间的,岗位年薪标准以区间上限为标准。

猎头合同范本

人才猎头服务协议 甲方: 乙方:尚才荟猎头 甲方因业务发展需要,特委托乙方以猎头服务方式招聘 _____________ 职位人选,甲乙双方 本着“平等合作,互惠互利”的原则,经友好协商达成如下协议: 一、甲方权利与义务 1.甲方应向乙方提供详细、真实的公司背景资料,包括但不限于公司背景、经营状况、发展战略、企业文化等。 2.甲方须负责提供所需招聘岗位的详细资料,真实填写《企业职务需求表》(附件1),并对薪酬、福利、休假等与应聘者利益相关一切信息之真实性负责。 3.收到乙方提交的候选人报告资料后,甲方应在三个工作日内通知乙方是否要求候选人面 试。如果面试,应提前告知时间和地点。面试结束后,甲方应在三个工作日内告知乙方面试结 果。对不合适的候选人,应给出说明。如果不面试,应给出具体的不面试原因。 4.甲方应在面试后积极与乙方沟通、协调,三天内做出上一轮面试的决议(包括安排下一 轮复试或录用)。特殊情况外,每个候选人面试次数最多为四次。 5.甲方对候选人信息、候选人报告、候选人面试过程等一切可能有损于候选人正当利益或 本合同正当履行的信息,负有保密义务。 二、乙方权利与义务 1.乙方对甲方提供的企业、职位等信息负有保密义务。除为完成本合同所必须外,其余未 经甲方授权,不得公开。 2.乙方需在寻访服务期(自—年—月—日至—年—月—日)内利用各种渠道为甲 方寻访、招聘所需人才。乙方在了解候选人的情况,分析其背景资料,并在此基础上进行筛选 后,向甲方提交合适候选人的个人资料。如果在寻访期内,甲方未能对乙方提供的候选人给出明确答

复,乙方有权终止合作。 3.若出现以下任何一种情况,均视为乙方已完成猎头服务全部工作,乙方有权要求甲方按本合同所列的“全职服务费用”,按约定支付周期与方式支付费用。如逾期支付,乙方将保留诉诸法律解决的权利: (1)甲方面试乙方所推荐的候选人后录用其作为甲方员工以全职合作; (2)甲方面试乙方所推荐的候选人后因某种原因不能提供所聘用岗位,但录取该候选人为其他岗位职员。 4.若出现以下任何一种情况,均视为乙方已完成猎头服务全部工作,乙方有权要求甲方按本合同所列的“全职服务费用”,在五个工作日内支付全额猎头服务费。如逾期支付,乙方将保留诉诸法律解决的权利: (1)甲方将乙方所推荐的候选人转荐给其他雇主,并且候选人被他雇主其聘用的; (2)甲方面试乙方所推荐的候选人后表示拒绝聘用,却在面试后一年内聘用乙方曾推荐的人选。 5.若出现以下任何一种情况,均视为乙方已完成猎头服务全部工作,乙方有权要求甲方按本合同所列的“兼职服务费用”,在五个工作日内支付全额猎头服务费。如逾期支付,乙方将保留诉诸法律解决的权利: (1)甲方面试乙方所推荐的候选人后因某种原因不能聘用该候选人,而聘用该候选人做兼职或其他短期服务; (2)甲方面试乙方所推荐的候选人后并没有聘用,而采用咨询、承包等方式与候选人保持合作关系。 三、服务费标准及支付方式 1.全职服务费用标准: □人民币 ____ 万元(大写:________ )(岗位年薪标准)X ____ % (年薪的百分比)□约定固定金额服务费用人民币 ______ 万元。 说明:支付方式按双方协商约定,二选一。 若甲方提出的岗位薪酬是某一区间的,岗位年薪标准以区间上限为标准。

猎头服务合同word排版格式

猎头服务合同word排版格式 甲方(劳务用工单位): 工商登记号:联系地址:邮政编码:联系电话:传真:联系人: 开户行:帐号: 乙方(劳务招聘单位): 工商登记号:联系地址:邮政编码:联系电话:传真:联系人: 开户行:帐号: 甲、乙双方就乙方向甲方提供需求职位人才的寻访推荐服务达成如下合同(以下称“本合同”):前言: 甲乙双方保证各自为依据中华人民共和国的相关法律法规合法成立的组织。乙方应确认有从事该项服务的资格,甲方应确认向乙方提供的各种资料的真实性,双方就各自的确认承担各自的法律责任。 甲、乙双方本着自愿和互惠互利的原则,合作签署“本合同”,并愿意自觉遵守“本合同”的各项条款。 第一条服务内容

1、甲方聘请乙方作为其中、高级人才推荐的供应商。 2、甲方如有招聘需求,应向乙方提供委托招聘职位信息,内容包括需求岗位职责描述、胜任条件、薪资福利条件、需求人数和日期等。 3、自甲方正式委托职位寻访之日起,乙方正式启动寻访推荐工作。15个工作日内向甲方推荐每个职位2—3名经过甄选基本符合甲方职位需求的候选人,直到甲方录用乙方所推荐的候选人为止。 4、乙方在推荐候选人时,需要向甲方提供人选的详细简历,说明其以往的工作经历、主要职责、成就以及能说明其工作胜任能力的任何其它必要资料。 5、甲方聘用乙方推荐的人员,在个月(保证期),自推荐人开始在甲方工作之日起,如发生聘用终止,乙方应为甲方免费提供一次相同职位的人才服务,乙方应保证在甲方通知乙方的15个工作日内为甲方再次提供候选人。 第二条限制 1、甲方不得提供虚假职位需求信息和做出虚假承诺。 2、合作期间甲方不得直接聘用乙方猎头顾问和工作人员,如有发生甲方应付乙方该职位应付猎头佣金的双倍的服务费用。 第三条保密原则 1、乙方应对甲方提供的任何商业、技术资料及员工信息进行保密。 2、甲方应对乙方提供的人选资料、服务费用保密,不得将乙方提供的人选资料用于任何非自己聘用内部员工的目的和提供给甲方之外的任何第三方(包括个人和法人)。 3、在未取得到推荐人选的允许,不得向其正在任职的单位(雇主)作背景调查。 第四条人选的推荐、面试 1、甲方在乙方提交候选人资料后的48小时内需将是否面试的决定通知乙方,并在面试后

猎头服务合同模板通用版

猎头服务合同模板通用版 Headhunting service contract template general version 甲方:___________________________ 乙方:___________________________ 签订日期:____ 年 ____ 月 ____ 日 合同编号:XX-2020-01

猎头服务合同模板通用版 前言:合同是民事主体之间设立、变更、终止民事法律关系的协议。依法成立的合同,受法律保护。本文档根据服务合同内容要求和特点展开说明,具有实践指导意义,便于学习和使用,本文档下载后内容可按需编辑修改及打印。 委托方:____________________(以下简称甲方)代理方:_________人才服务部(以下简称乙方)甲方因业务发展 需要,委托乙方搜索所需人才。经双方友好协商,达成以下合作条款: 第一条主旨本合同目的在于确定乙方代表甲方搜 索候选人时双方的权利及义务。 第二条搜索服务的基本方式 1.甲方需提供与所需搜索职位有关的详细资料给乙方, 其中主要包括:--工作环境,如公司背景、现状(规模、发展状况、目前和将来的投资等);--职务名称、职位职责,个人发展前景、直属上司情况及汇报路线;--所须具备的条件(包括年龄、性别、教育程度、英文程度、专业经验以及个人品性);--工作中所需要的独立性、工作氛围,流动性(出差需要)和上班地点等;--福利待遇(包括基本工资、奖金和津贴、工作时数、生活条件以及合同保险)。 2.乙方工作程序:--搜索可能达到该标准的候选人,

对这些候选人进行初次面试筛选并确证其资料的真实性; --将候选人资料提交甲方参考,包括对其经验的评价及对其性格,能力和潜质的看法;--为双方安排面谈时间、地点,并参与其中一些条款的协商,协助聘用合同的最终签署; --协助甲方督促被录用候选人与原工作单位按正常的程序办好离职手续。 3.乙方在合同签定之日起天内完成对甲方所需人才的寻访、测评及背景调查,将完整、真实的候选人资料提交甲方,并安排候选人面试。 4.甲方应在收到乙方提供的候选人资料后两日内,做出是否与己有人才资料重复的判断,否则视为乙方推荐;甲方应在收到乙方提供的候选人才资料后一周内,做出是否需要复试的判断,甲方应在候选取人面试后两周内将面试意见及审核结果反馈给乙方,在复试后四周内做出是否录用的判断。 第三条保证期 1.推荐的候选人被甲方录用,并最终到甲方工作,视为猎头服务成功。 2.雇员在保证期(三个月)内无论因任何原因离职,甲方应五天内向乙方提出书面报告,乙方将免费为甲方推荐另一合适候选人到职;乙方收到甲方书面报告后二个月内没有推荐合格人选到职,乙方应退还该项服务费的_______%。

猎头服务合同模板公司方修订版

人才访寻服务合同 本合同由以下双方于年月日在北京签署: 甲方: 地址: 电话: 联系人: 工作邮箱: 乙方: 地址: 电话: 联系人: 工作邮箱: 甲方与乙方经友好协商,双方同意并签订以下条款: 一、服务 1.甲方非独家委托乙方在中华人民共和国境内(不包括港、澳、台地区)按照甲方的职位要求,检索、筛选合适的人选推荐给甲方。为甲方搜索和审核核实候选人,并进行必要的面试与评价。同时,根据甲方在委托期间对招聘计划进行的调整作积极配合及给予相应更换。2.此服务以非专有的方式提供,即:甲方有权聘请任何第三方提供此项服务,乙方有权向任何第三方提供类似之服务,若第三方的职位要求与甲方类似,乙方应优先向甲方推荐合适人选。 3.向被聘用的候选人提供人事咨询与帮助,与甲方保持不间断的关系,协助双方解决可能遇到的问题,确保候选人顺利入职。 4. 甲方与被聘用人员之间的聘用关系由甲方与被聘用人员负责。乙方不对甲方与被聘用人员之 间、任何其他第三方(个人或法人)之间的关系负责。 二、职位描述 1.甲方将随时给予乙方空缺职位描述,乙方应在收到空缺职位描述后1个工作日内与甲方沟通相关职位需求。 2、如乙方为寻访推荐候选人进行必要的广告宣传,应依法、诚信、真实宣传,不得损害甲方的 合法权益,发生的相关费用均由乙方自行承担,乙方对外宣传的广告信息或资料及其所有对外发布信息内容须经甲方书面认可后发布。如果因为广告宣传损害甲方或者第三方合法权益的,由乙方自行承担责任。 三、人员推荐

1.乙方应当按照甲方的招聘需求面试候选人、出具面试报告、推荐合适的人选,并应在接到甲方的招聘需求后5个工作日内就每个职位推荐至少3名符合甲方职位要求的候选人,直到甲方聘用乙方所推荐的候选人为止。 2.乙方应根据甲方的要求核验候选人的个人资料,并做背景调查,说明其以往的工作经历、主要职责、成就以及能说明其工作胜任能力的任何必要资料,保证合法的向甲方提供的每一位候选人的资料真实、准确、完整。

猎头服务协议书

猎头服务协议书 甲方:____________服务有限公司 ____________人才服务中心 乙方:________________________ 甲乙双方就甲方向乙方提供人才猎头服务,达成如下协议: 一、根据中华人民共和国的法律和法规,甲方确认本方有完全的从事本协议所述服务的资格;乙方确认本方是依法成立的组织,符合法律规定的面向社会公开招聘的各项条件和手续,并且向甲方提交的各种书面文件及陈述都是真实有效的。甲乙双方就各自的确认承担相应的法律责任。 二、甲方服务的收费标准是,每成功提供一人,收取________元人民币作为本方的服务佣金。佣金分两次支付,双方签定协议之日,甲方支付________元/人为预付佣金;推荐人选上岗三个月或转正后一个星期内付清余款。 三、甲方收到乙方预付的预付服务佣金后,应按照本协议附件1的要求为乙方提供人才,并保证所提供人才情况的真实性,同时协助乙方安排有关面试考核的事宜。面试考核应本着方便应聘者及乙方的原则进行。 四、乙方应按照本协议附件1规定的条件考核验收甲方提供的人才。并应于考核之日起________天内将是否聘用(含试用,下同)的意见及理由书面通知甲方。否则,视为同意聘用。

五、甲方应于收到预付服务佣金之日起两个星期至一个月内完成委托事项,经双方协商同意,上述期限可以改变,但应另以书面文件规定之。 六、乙方同意聘用(接纳为乙方正式员工)甲方提供的人才,甲方服务即告完成。乙方若于聘用后三个月内辞退甲方提供的人才,应于辞退之前通知甲方,并有权要求甲方重新提供服务,甲方应免费为乙方再推荐人选,并得到乙方聘用为止。 七、非因不可抗力,或乙方认可的其他正当理由,甲方在协议规定的期限内未能向乙方提供候选人才的,应全额返还乙方预付佣金;虽提供候选人才而乙方最终未能聘用的,预付佣金应退还甲方。 八、乙方以超出本协议附件1以外的条件拒绝甲方提供的人才,或在本协议规定的有效期限内自行招聘协议附件1规定的人才,从而无需甲方的服务,甲方有权留置乙方全部预付佣金,折抵甲方服务收入。 九、乙方拒绝甲方提供的人才,从而导致甲方服务失败,但在甲方服务终止后的十二个月内聘用曾被乙方拒绝的人才,应于聘用前通知甲方,并向甲方支付全额服务佣金。否则视为违约,乙方应就此向甲方支付________至________倍的服务佣金作为违约金。 十、本协议在乙方委托寻猎的人数超过一人时,对每一个人才的服务分别有效,彼此无法律上的连带责任关系。

相关主题