A Novel Type-2 Fuzzy Directed Hybrid Post-Filtering Technique for Efficient JPEG Image Artifact Reduction

摘要

Image and video compression becomes very popular and intense field in recent decades, due to their fast and quality communication demand. To cope up with the high speed communication demand of image and videos, over the communication channel, higher compression (ie. low bit rate compression) is now preferred at the cost of quality degradation. Currently JPEG and JPEG2000 are the most popularly used codec's for achieving quality image compression. Practically, on the low bit rate compression, it has been observed that, the JPEG standard compressed images suffer from multifarious visual distortions. To address this problem effectively, this paper proposed an innovative type-2 fuzzy directed hybrid post filtering technique, which is framed to suppress the artifacts generated in JPEG compressed images at low bit rate compression. The proposed technique addresses all three types of JPEG compressed image artifacts: blocking, edges blurring, and aliasing. Furthermore the proposed technique is structured with two stages, to enhance the quality of JPEG compressed images. The first stage, removes blocking artifacts using boundary smoothing and guided filtering techniques. The second stage reduces blurring and aliasing around the edges through type-2 fuzzy directed local edge regeneration. To prove higher efficiency of proposed work, a complete comparative performance evaluation with other existing JPEG artifact removal techniques, has also presented. This extensive performance analysis includes visual quality assessment of post-filtered images along with subjective quality assessment on the basis of, Peak Signal to Noise Ratio (PSNR) and Mean Square Error (MSE). Performance evaluation illustrates that the, proposed approach is efficient, provides PSNR improvement of approximately 1 dB along with higher reduction in MSE values as compared to state of the art algorithms for all the bit-rate compressions.