Smart removal of blocking effects from video

Contact person: Dmitriy S. Vatolin (dmitriy@graphics.cs.msu.ru)

MSU Graphics & Media Lab (Video Group)

The filter is intended for high quality recovering visual quality of video compressed using conventional block-transform-based codecs (for example, video ripped from DVD, VideoCD or after decompressing by H.261, H.263, H.264, DivX, XviD, etc).
The main advantage of developed method is hard blockiness suppression together with preserving of objects' details. Hard blockiness suppression is applied in case of extremely low bitrate when entirely flat single color blocks are appeared on the frame. It is performed independently in the first phase of the algorithm and can be disabled manually.

Moreover, this method uses technique of automatical blockiness level detection which is also used in fast deblocking algorithm. Using this technique algorithm automatically determines the blockiness's strength on the frame and in the local areas, that allows to save the maximum of details. Thus, method will blur marginally when on scenes with slow motion, while it will smooth very hard in the case of fast motion in the same video.

Examples

Samples below illustrate filter's performance.

Hard blockiness suppression feature performance is showed on the folowing pictures (frame fragment of sequence "battle" - the part of "Terminator II" movie). The middle picture illustrates the result of processing the left picture using conventional deblocking method. Right picture shows the result of processing the left picture using hard blockiness suppression feature. It is obvious that right sample demonstrate significantly higher visual quality in terms of blockiness suppression.

Decompressed frame After conventional deblocking After hard blockiness suppression feature

Pictures below illustrates visual performance of deblocking algorithm in case of hard blockiness. Focus your special attention on processing of single color flat blocks:

Frame with blockiness Frame after Smart Deblocking

Pictures below illustrates visual performance of deblocking algorithm in case of presence of higly textured area:

Frame with blockiness Frame after Smart Deblocking
Frame with blockiness Frame after Smart Deblocking

The next example demonstrates performance results for several deblocking methods. During analysis focus your special attention in the area near the center of the ball:

Frame with blockiness MPlayer Deblocking
University of Dallas deblocking Proposed algortihm

The diagram below demonstrates objective comparison results. Vertical axis is marked by average Y-PSNR gain for sequence. Y-PSNR gain is calculated as the difference between Y-PSNR value of compressed sequence processed by deblocking and compressed sequence without any processing. Fast deblocking (see link at the bottom of current page) demonstrates the best objective result but it produces lower visual quality comparing to Smart Deblocking:

Publications

D.Vatolin, S.Grishin, "Method for High Quality Deblocking of Video without use of Quantization Information", Graphicon-2004. (Russian)

Download

Filter for VirtualDub MSU SmartDeblocking (355KB, ZIP)
Unpack this file in the folder "Plugins" in VirtualDub directory to use.

See also

Fast deblocking algorithm
Detailed description of MSU SmartDeblocking filter
MSU Deblocking filter page (other our deblocking algorithm)

Source frame Bicubic-resized frame Processed with MSU Video2Photo frame
Bicubic-resized frame Processed with MSU Video2Photo frame


Source frame

Team