DCT-like transform for image compression requires 14 additions only

Abstract

A low-complexity 8-point orthogonal approximate DCT is introduced. The proposed transform requires no multiplications or bit-shift operations. The derived fast algorithm requires only 14 additions, less than any existing DCT approximation. Moreover, in several image compression scenarios, the proposed transform could outperform the well-known signed DCT, as well as state-of-the-art algorithms.

Comments on ‘Area and power efficient DCT architecture for image compression’ by Dhandapani and Ramachandran

“…This seems contradictory. However, to examine the transformation suggested by the authors, we adopted a constant bitrate encoding based on the retention of r transform-domain coefficients, as suggested originally by Haweel and others [2][3][4][5][6][7][8][9][10][11].…”

“…Thus, we considered a Monte Carlo simulation with 10,000 8-point input vectors modeled after the firstorder Markov process with correlation coefficient of 0.95 [12]. For comparison, we considered the following transformations: the DCT [12], the SDCT [2] the BAS-2013 [9], and the BC-2012 [3]. Obtained mean values are displayed in Fig.…”

“…To further assess the claim of good coding capabilities, we considered the unified coding gain and the transform efficiency as measures to quantify the coding performance [12] of T in comparison with bona fide transforms, such as: DCT, SDCT [2], BAS-2008 [4], BAS-2009 [6], BAS-2010 [8], BAS-2011 [9], CB-2011 [13], and CB-2012 [3]. In addition, we also considered the transformation in [14] and transformation in [15].…”

Area and Power Efficient VLSI Architecture for Two Dimensional 16-point Modified Gate Diffusion Input Discrete Cosine Transform

“…In [5], a low complexity approximate was introduced by Bouguezel et In [8], CB-2011 is modified by replacing elements of the CB-2011 matrix with zeros. The transformation matrix C 1 =D 1 .T 1 where D 1 =diag (1/√8, 1/√2, 1/2, 1/√2, 1/√8, 1/√2, 1/2, 1/√2}.It needs only 14 addition for its computation.…”

“…The proposed digital architectures are custom designed for the real-time implementation of the approximate algorithms [5][6][7][8][9][10][11].The proposed architecture employs two parallel realizations of DCT approximation blocks, as shown in Fig. 5.…”