Deinterlacing using Motion Detection SANTHOSH KUMAR K.C.
Motion Detection in Deinterlacing Motion Detection is used in Deinterlacing process Interlaced Scanning was used to lower the costs of video system and reduce the transmission bandwidth by 2 Deinterlacing is translation of the interlace format of video to progressive device as plasma projection TV, projector, and computer Difficult to police: very large and distributed user base – Trust network: clusters of users sharing the same social interests developing trust with each other – Platform openness for developing applications that are attractive the general users who will install them
Deinterlacing The challenge of deinterlacing is to interpolate the missing points with limited information and also to maintain clear visual quality as well The key concept of deinterlacing is to interpolate the missing point with neighbours that have the highest correlation
Earlier methods employed LA-Line Average –intra interpolation-average to two pixels on the same line. But LA blurs the vertical details. FI-Field Insertion - Inter interpolation- repeats the pixel p(i,j-1) as p(i,j) – but caused problem with the moving objects in the frame to be deinterlaced. So many Motion adaptive algorithms were designed. But the visual quality of these algorithms depended on the correctness of motion detection which needed more field memory in VLSI implementation
Hybrid Motion Deinterlacing Algorithm This algorithm consisted of hybrid Motion Detection(HMD) and edge Pattern Recognition(EPR). HMD for detection of versatile motion EPR for interpolation of edges and textures which cannot be handled by LA and enhanced LA Algorithms
Block Diagram - HMDEPR from paper Motion-Adaptive Deinterlacer via Hybrid Motion Detection and Edge –Pattern Recognition by Gwo giun Lee ,Ming-Jiun Wang,Hsin-Te Li, and He-Yuan Lin
HMD-Hybrid Motion Detection
Pseudocode from paper Motion-Adaptive Deinterlacer via Hybrid Motion Detection and Edge –Pattern Recognition by Gwo giun Lee ,Ming-Jiun Wang,Hsin-Te Li, and He-Yuan Lin
HMD-Continued First Condition – diff(a-b) > TH1 This condition detects slow motions in the frame Second Condition – diff(b-(c+d)/2) > TH1 and diff(b- (g+h)/2) < TH2 This condition detects the fast motion in the frame Third Condition – abs(a+(e+f)/2 – b- (c+d)/2) > TH3 This condition detects motion with edges
Erosion after HMD Erosion filter with cross shaped mask is performed on the results of HMD. The erosion filter is to eliminate the isolated moving pixels
Erosion Mask
Dilation A dilation filter with a 3x3 Mask is performed after erosion to restore and extend the shape of moving objects after erosion.
Dilation Mask
Corrects the Motion missing.. The inability to detect motion is called as motion missing which results in motion holes The earlier algorithms had these problems The Erosion and Dilation process after HMD eliminates this motion missing problem and hence gives good visual quality.
Interpolation follows HMD EPR-Edge Pattern Recognizer -The intra field interpolator used in moving scenes FI – Field Insertion - The inter field interpolator used in stationary scenes.
EPR – Edge Pattern Recognizer
Experiental results of this algorithm
Summary of this algorithm The Hybrid Motion Detector is capable of detecting slow motion, fast motion and motion of edges with high accuracy. Edge pattern Recognition algorithm performs adaptive interpolation and thus achieves successful interpolation of textures and edges.
Hierarchal Motion Detection Algorithm First step is measuring the motion activity level in video sequence and calculating the adaptation factor Interpolation is spatial or temporal depending on the adaptation factor Spatial interpolation for static sequence and temporal interpolation for dynamic motion sequence
Block Diagram A Motion Adaptive Deinterlacing method with hierarchical motion detection algorithm by Ellan shahinfard,Maher A.Sid –Ahmed, Majid Ahmadi
Motion Detection Uses five consecutive fields of video data to increase the ability of detecting fast motions. In the fig(next slide),F represents the current field,F+1 and F+2 represents the two subsequent fields, and F-1 and F-2 represent the two fields prior to F.
Block Diagram A Motion Adaptive Deinterlacing method with hierarchical motion detection algorithm by Ellan shahinfard,Maher A.Sid –Ahmed, Majid Ahmadi
(F-2,F),(F-1,F+1),(F,F+2) have same missing lines and same time difference. The three pairs are used for motion detection The Maximum of three corresponding diff block is compared with predefined threshold value. If less than threshold value, then static video sequence Else then dynamic block is recursively partitioned to smaller blocks and procedure continues.
Interpolation Depending on the mobility value calculated in the previous motion detection, spatial interpolation is chosen for static sequence or temporal sequence is chosen for dynamic sequence.
Experimental results
A Motion Adaptive Deinterlacing method with hierarchical motion detection algorithm by Ellan shahinfard,Maher A.Sid –Ahmed, Majid Ahmadi
References A Motion Adaptive Deinterlacing Method with hierarchical motion detection algorithm – Elham Shahinfard,Maher A.sid-Ahmed,Majid Ahmadi A Motion-Adaptive Deinterlacer via Hybrid Motion Detection and Edge-Pattern Recognition - Gwo Giun Lee,Ming-Jiun Wang,Hsin-Te Li and He-Yuan Lin. An improved Motion Adaptive Deinterlacing Method using variable block size Motion Detection - Elham Shahinfard,Maher Sid-Ahmed,Majid Ahmadi