No abstract

No abstract

In this paper, a novel error concealment technique based on the discrete cosine transform ( D m coefficients recovery is presented, and the estimating method of DC and low frequency coefficients is described in detail. The proposed technique makes full use of the neighboring information around the missing blocks, instead of only considering a row of pixels. The proposed algorithm is simple and is applicable to any unitary block-transform and is very effective for recovering the DC and low frequency coefficients. Simulation results show that the quality of recovered images is significantly improved in different marcoblock loss (MB) rate. ABSTRACTIn this paper, a model-based algorithm is described to segment the pectoral muscle in the mammograms. Afier the mammogram has been preprocessed, a set of ROI with different sizes is applied on it, in each of which, an iterative thresholding technique is used to gain an optimal threshold correspondingly. All the thresholds compose the threshold curve and e curve, which the optimal threshold of the mammogram could be extracted by. Finally, a twice line fitting and polygon approaching technique is carried out to refine and approach the edge curve of the pectoral muscle, which has been partly segmented by the thresholding. About 60 mammograms have been used to evaluate the proposed algorithm with satisfactory result reached.

n e MPEG-2 compression algorithm is very sensitive lo channel disturbances due to the use ofvariable-length coding. A single bit error during transmission leads to noticeable degradation ofthe decoded sequence qual@, in that pa^ or an entire slice infirmation is lost until the next resynchronizntion point is reached. E m r concealment (EC) methods, implemented at the decoded side, present one way of dealing with this problem. In this paper, based on smoothness, consistency, andfractal behavior, A spatial domain ermr concealment algorithm with texture recovery for MPEG-2 video streams is presented by using rhe full information of a4acent macroblocks (ME)), instead of using only a onepirel wide area around the missing block Since the left and right neighboring MBs ofthe lost ME are often erroneous, the proposed algorithm utilize only the top and bottom MBs ofthe lostMB Io conceal the lost data.The experimental results show that the proposed algorithm can recovery texture information in lost MBs andprovide better visual qua& in comparison with the conventional spatial domain inteiplation algorithms.

No abstract