Communication Technology, Vol 3, Issue, January- 4 ISS (Print) 23-556 Key Frame xtraction and Shot Change Detection for compressing Color Video Dr. A. SKhobragade, eha S Wahab Dept.of &T ngineering YeshwantraoChavan College of ngineering,agpur,india atish_khogragade@rediffmail.com, nehashammiwahab@gmail.com Abstract Progress intechnology is rapidly increasing now-a-days. The number of user are increasing but the resources are limited. So this project is to effectively and efficiently use the limited and expensive resource, soas to cater every individual requirement. The color video will be compressed with the help of two algorithm: Shot Change Detection and Key Frame xtraction. Broadly video is into segmented into scene, shot and frame respectively. Application of this project key is video broadcasting, video streaming, handled device, portable multimedia player. Index Terms color, compression, scene, video broadcasting. I. ITRODUCTIO: There is an exponential increase in use of digital media,sothere is a strong need for finding out new methods for efficient storage of video is essential. A Color Video occupies more space than a gray scale video. If a color video is converted with partially colored video significant increase compression is possible. Partially colored video consist of Key Frames For efficient management of video data, shot change detection( SCD) technique is used. Shot change detection (SCD) is a algorithm or a procedure for detecting changing frame in video and one of the important techniques required for efficient management of video date.scd algorithm have been studied largely in two domain; spatial and frequency.zhang et.al have introduce compression for each pixel between adjacent two frames. However,the drawback of this method is failing to distinguish large variation in a small area of pixels or slight variation in a large area of video data. Also heproposed a block based scheme using the likelihoods ratio compression and features of local area of images to reduce sensitive to movement of the camera and object. This method determines shot change frame with variation of corresponding blocks of frames. Shot change in frame is widely used in Histogram based method. A. Threshold setting With help of eye,we can make decision where in a video is a shot change,but in eye concept should be replaced with some comparison value to detect short change in video data. This value is called a threshold. Threshold is most important II SHOT CHAG DTCTIO Figure :Block Diagram of project. www.ijrcct.org Page 67
Communication Technology, Vol 3, Issue, January- 4 ISS (Print) 23-556 element is SCD algorithm because it is a decision factor. The SCD algorithm method if of two types : the fixed threshold method and adaptive threshold method.from repeated experiments,the fixed threshold method determines optimal threshold. But they require much experiment iteration and must find other optimal threshold for other video data. The adaptive based threshold SCD algorithm gets suboptimal threshold according to devised rules since it is not essential to find threshold manually for each video with a For adaptive threshold, we use meanand variance of the consecutive frames. qs. () presents the mean value of frame difference..qs.(2) presents the variance of mean value (2) () In the equations, m denotes the means of frame difference and v means variance next frame f i is a current frame and f i+ the subsequent frame of f i.w and H are horizontal and vertical size of frame. (4) (3) qs. (3) andqs.(4) are weighting variance can be calculated by dividing the scaled W and H. As shown in qs. () and (3),m is calculated with W/wd and H/wd instead of Wand H. Similarly, we can calculate v with m, W/wd and H/wd instead of W and H from q. (2). B. Setting adaptive thresholds For determination of threshold for a shot change frame is an important element in procedure of SCD. In general if the frames of a video are larger than a given threshold we regard frame as shot change. Thus, we determine a video frame a as shot change frame. SCD method for fixed threshold have been studied extensively. They detect shot change frames with some fixed value which are called threshold. Repeated iterationfor adjustment of thresholds is done until they get the best result. In general, variation of threshold is relatively large to find a fixed threshold.. Hence because of these variation of thresholds,adaptive thresholds are required which are calculated with few frames. III.Key Frame Selection Due to rapid increase in amount of video data generated and wide range of video application,an efficient and effective management of data is need of hour. A video consist of group of frames (GOF).To manage the video data, key frame selection is important. Generally key frame are I,P,B frame. I is the independent or Intra frame. P is the Predictive frame or non-independent frame.b is the bidirectional frame.it moves either in forward or backward direction. Some traditional method for frame selection is the first, middle, last frame. Generally for a video sequence with low motion activity very few frames(about.5% of all frames) are selected with higher compression.for a video sequence with higher motion activity, more number of key frames(.5%of all frames ) are selected. B. Temporally Maximum Occurrence Frame(TMOF) The optimal key frame is constructed by considering the probability of occurrence of those pixel at corresponding along the frames in a video. This constructed key frame is TMOF.For further optimization there are two scheme namely k- TMOF and kp-tmof. k-tmof pixel value with largest probability of occurrence are selected.in kp- TMOF highest peak of probability distribution of occurrence at each each pixel position for a video www.ijrcct.org Page 68
Communication Technology, Vol 3, Issue, January- 4 ISS (Print) 23-556 is considered.based on this TMOF, considering it as reference frame distance of each frame is calculated. Then after averaging every distance,a threshold is found. And then individually comparing, their respective distance with this threshold,one can decide whether the frame is key frame or not. Frame # IV. XPRIMTAL RSULT The video chosen is vipmen.it consist of total 238 frame.for pixel wise comparison the frame is divided in block like four quadrant. a b c d TABL :VIDO RSULT AM VALU VIDO VIPM FRAM SIZ(W*H) * Figure 2: Individual frame divided into 4 blocks.here frame no and frame no are divided into 4 blocks..5 x 0-3 UWIGHTD VARIAC of FRAMS KY FRAMS SHOT DTCTIO FRAM Frame # C IA R A V D T H IG W U 0.5 0-50 0 50 50 0 250 FRAMS 7 x 06 6 Fig- Frames WIGHTD VARIAC of FRAMS a b 5 C IA 4 R A V D 3 T H I G W 2 c d 0-50 0 50 50 0 250 FRAMS www.ijrcct.org Page 69
Communication Technology, Vol 3, Issue, January- 4 ISS (Print) 23-556 Figure 3: Histogram of unweighted variance and then weighted variance by a factor of 4. FRAM O:3 FRAM O: FRAM O:7 FRAM O:4 FRAM O:9 FRAM O:0 FRAM O:2 FRAM O:7 FRAM O:22 FRAM O: Figure 4: lven () KY Frame from video vipmen consisting 238 frames. www.ijrcct.org Page 70
Communication Technology, Vol 3, Issue, January- 4 ISS (Print) 23-556 V.COCLUSIO In this project,with using shot detection and key frame extraction compression of video is achieved. Key frame or visually important frame are selected discarding other frames in a video to achieve compression. The reason for this compression is just look only the key frame and not the entire video and also saving time. Shot detection algorithm is applied to video,gets one shot which implies there gradual change in information in consecutive frames. RFRCS [].R.Agrawal,S.Gupta,VGude, A Color video compression technique using key frames and a low complexity color transfer.international Conference on Signal Processing and Communication (SPCOM)July 25.2. [2]Won Hee Kim, Jong amkim, An adaptive shot change detection algorithm and its implementation on portable multimedia player. I Transaction on Consumers lectronics Vol55.o.2,May 09. [3].Sze,KW,Lam K,M and Qiu,G, Anew key frame representation for video segment retrieval I TransCircuitsandSystVideoTechnology,Vol5.o. 9,Sept 05. www.ijrcct.org Page 7