Method and device for compressing an image signal

A device for compressing a digital image signal, comprising:

• means of inputting data represented by coefficients allocated to the pixels of an image,
• means of forming spatial blocks of N×M coefficients allocated to the pixels of the image, where N and M are non-zero integers,
• means of transforming each spatial block in a frequency domain in order to form a frequency block of N×M coefficients, and
• means of processing image segments, each one consisting of a given number of macroblocks, each consisting of a given number of frequency blocks, in which the means of processing a segment comprise:
• first means of quantizing the coefficients of each frequency block of the segment, in order to select a first quantization step set, attached to each frequency block and defined for each frequency block by an index group including an index which is identical for all the frequency blocks of one macroblock,
• second quantization means for quantizing at least some of the frequency blocks, with a second quantization step set substituted for the first quantization step set, and
• means of calculating, for each block, a gain factor representing a reduction, on passing from one quantization step set to the other, of the amount of data needed to represent the quantized coefficients, this gain factor being taken at the level of the second quantization means in order to select the frequency blocks for which the first quantization step set is substituted by the second quantization step set.

A program which can be loaded into a memorycomprising portions of codes to implement a method of compressing a digital image signal having coefficients allocated to the pixels of an image, on execution of said program by a processor performs the following:

• forming spatial blocks of N×M coefficients allocated to the pixels of the image, where N and M are non-zero integers;
• transforming each spatial block in a frequency domain in order to form a frequency block of N×M coefficients;
• and
• grouping a plurality of frequency blocks in macroblocks and of a plurality of macroblocks in segments, together with the following steps for each segment:
• first quantization of the coefficients of each frequency block, with a first quantization step set, defined for each frequency block by an index group including an index which is identical for all the frequency blocks of a macroblock, and
• second quantization of the coefficients of at least some of the frequency blocks, with a second quantization step set substituted for the first quantization step set,
• the first and the second quantization step sets being chosen according, on the one hand, to a target value of the amount of data needed to represent the quantized coefficients of the segment and, on the other hand, to gain factors respectively calculated for the frequency blocks of the segment, the gain factor relating to one block representing a reduction, on passing from one quantization step set to the other for the said block, of the amount of data necessary to represent the quantized coefficients.

A method of compressing a digital image signal having coefficients allocated to the pixels of an image, the method comprising:

• forming spatial blocks of N×M coefficients allocated to the pixels of the image, where N and M are non-zero integers;
• transforming each spatial block in a frequency domain in order to form a frequency block of N×M coefficients;
• grouping a plurality of frequency blocks in macroblocks and of a plurality of macroblocks in segments by defining an index group for the frequency blocks of a macro block;
• quantizing the coefficients of each frequency block, with a first quantization step set, defined for each frequency block by an index group including an index which is identical for all the frequency blocks of a macroblock, and quantizing the coefficients of at least some of the frequency blocks, with a second quantization step set substituted for the first quantization step set.