Two‐layer motion estimation algorithm for video coding

Abstract

A novel two-layer motion estimation which searches motion vectors on two layers with partial distortion measures in order to reduce the overwhelming computational complexity of motion estimation (ME) in video coding is proposed. A layer is an image which is derived from the reference frame such that the summation of a block of pixels in the reference frame determines the point of a layer. It has been noted on different video sequences that many motion vectors on the layers are the same as those searched on the reference frame. Experimental results on a wide variety of video sequences show that the proposed algorithm achieves both fast speed and good motion prediction quality when compared with the state-of-the-art fast block matching algorithms.Introduction: Most of the fast search motion estimation (ME) algorithms endeavour to reduce the computational cost of ME greatly by checking only a few search points inside the search area by using the full distortion measure. Some of these algorithms [1-4] have reduced as many search points as possible to reduce the computational cost of ME. It is very difficult to reduce the number of search points further without degrading motion prediction quality. The computational cost of ME can be cut down not only by reducing the number of search points, but also diminishing the number of computations in the block distortion measure. This implies a partial distortion measure instead of the full distortion measure used at each search point. In this Letter, a two-layer ME (TME) is proposed. This algorithm first constructs the layers from the reference frame so as to facilitate the calculation of partial distortion measures on the layers with a lesser number of computations. Later, it performs ME by computing the partial distortion measures on the layer.

High Precision of Global Motion Estimation in Multi-dimensional Transform Domain

“…The special energy footprint in frequency domain of MVM-DCT coefficients has been proved in reference [4], that is to say, the data of large coefficients are concentrated around a folded plane. As the plane is always passing the origin of the frequency domain and is orthogonal with the direction of motion vector, so by detecting and locating the plane, we can .…”

Derin Öğrenme ile Görüntülerde Gürültü Azaltma Üzerine Kapsamlı Bir İnceleme

“…Son olarak elde edilen görüntüye gürültü gideren evrişimsel sinir ağı yöntemi uygulanmaktadır. DVDNet yöntemi geliştirilip U-ağı uygulanarak FastDVDNet yöntemi geliştirilmiştir [141]. FITVNet yöntemi iki aşamalı bir gürültü giderme modeline sahiptir.…”

Fast motion estimation algorithm using multilevel distortion search in Walsh–Hadamard domain

“…However, it is difficult to ensure the accuracy of prediction which may cause degradation dramatically in both ME matching quality. The third category is the elimination search algorithms which only make use of a part of pixels in distortion computation so as to reduce the total computational complexity of ME, such as partial distortion search (PDS) [11–14] algorithm, successive elimination algorithm (SEA) [15], multilevel successive elimination algorithm (MSEA) [16, 17], transform‐domain successive elimination algorithm (TSEA) [18] and two‐layer motion estimation [19] algorithm. The PDS algorithm reduces the computational complexity by reducing the amount of calculation of block matching distortion.…”