Fast partial difference elimination algorithm based on block matching error prediction

Abstract

Abstract. We introduce a fast partial difference elimination ͑PDE͒ algorithm for motion estimation ͑ME͒. The basic idea of the proposed approach is to eliminate invalid candidates earlier by predicting a total matching error between matching and candidate blocks. The matching error prediction is performed by using a partially computed matching error between the blocks. Experimental results show that the proposed algorithm can significantly save about 40% of the averaged computational costs versus the conventional PDE algorithm for ME at the cost of ignorable image quality degradation, whose average value is 0.0012 dB.

A content adaptive fast PDE algorithm for motion estimation based on matching error prediction

“…For this purpose, we proposed a prediction scheme of the total block matching error bSAD from the partial SAD pSAD k by using Taylor series expansion [8]. The prediction scheme is based on the relationship between a partial block SAD and its total block SAD that the partial SAD has a random value and is gradually increased to the total block SAD as the computation goes [8], [10]. First, we estimated a partial SAD pSAD n at the nth line by using the partial SAD pSAD k accumulated up to the kth line of original and candidate blocks as follows:…”

“…It is obvious that predetermination of the elimination will help in reducing the computations and in accelerating the speed of ME. For this purpose, we proposed a prediction scheme of the total block matching error bSAD from the partial SAD pSAD k by using Taylor series expansion [8]. The prediction scheme is based on the relationship between a partial block SAD and its total block SAD that the partial SAD has a random value and is gradually increased to the total block SAD as the computation goes [8], [10].…”

“…However, the improved PDE approaches still requires heavy load for SAD computations and time to eliminate a candidate block at the final stage if it is very similar to a original block but not the optimal block. This paper is based on a previous work [8] of a predictionbased fast motion estimation. However, we employ two new main factors which are a sub-sampling based error prediction and a selective pixel-based error comparison for improving the prediction accuracy but reducing the computational cost, respectively.…”

“…Recently, a prediction-based algorithm has been introduced [8] and reported that it could save the computation cost considerably. However, it is not guaran- teed that the prediction is correct because a prediction at a candidate block depends on its neighboring blocks instead of its own contents.…”

Efficient partial distortion search algorithm for block based motion estimation

“…A two stage sorting based PDS algorithm by using the characteristic of pattern similarity matching error is described in [7,8]. The basic idea in [9] is to eliminate invalid candidates earlier by predicting a total matching error between matching and candidate block. However this prediction is based on the linear model which results significant degradation of image quality.…”