Video deblocking with fine-grained scalable complexity for embedded mobile computing

This paper addresses the need of reducing blocking artifacts after video decompression in embedded mobile computing devices such as mobile phones and PDAs with limited computational capability, where low bit rate coding is usually employed and video deblocking is highly desirable. A novel video deblocking method has been developed which consists of two steps: deblocking mode decision and deblock filtering. Blocking artifacts are detected by examining the value of several adjacent pixels. Depending on the degree of blocking artifacts, a filter mode and a corresponding filtering center are determined for a region of pixels. The deblocking filter is chosen from five different types of candidates including variable center filters and nonsymmetric filters. Extensive experiments show that the proposed algorithm has achieved both lower computational complexity and better visual quality as compared to MPEG-4 VM. Furthermore, targeting the need of embedded mobile computing platforms, a scheme is developed to dynamically scale the complexity (and hence power consumption) of the deblocking algorithm with graceful visual quality degradation.