CN107257473B - An Efficient Image Compression Algorithm - Google Patents

Abstract

The invention discloses an efficient image compression algorithm, which comprises the following steps: input: image; output: compressed bit stream. Step 1: Perform wavelet transformation on the image to generate a set of wavelet coefficients in a pyramid structure; Step 2, perform bit-level coding on the image after wavelet transformation. The compression speed of the algorithm of the present invention is comparable to that of the SPIHT algorithm, but the image quality PSNR under the same compression ratio is 0.2 to 0.4 dB higher than that of the SPIHT. The algorithm has image compression efficiency comparable to SPIHTwAAC algorithm and JPEG2000 algorithm with relatively low complexity. The invention has the characteristics of relatively low complexity and high compression rate, and is suitable for popularization and application.

Description

An Efficient Image Compression Algorithm

technical field

The present invention relates to an efficient image compression algorithm, in particular to an image compression algorithm with relatively low complexity and high compression rate.

Background technique

Image compression algorithm has a wide range of social application value. There are two main types of image compression algorithms at present: compression algorithms based on discrete Fourier transform (such as JPEG), and compression algorithms based on discrete wavelet changes (such as SPIHTwAAC, JPEG2000). The compression algorithm based on discrete Fourier transform has low complexity, fast compression rate, but low compression efficiency. The compression algorithm based on discrete wavelet transform has high compression efficiency, but high complexity.

SUMMARY OF THE INVENTION

The purpose of the present invention is to overcome the defects existing in the prior art, and provide an efficient image compression algorithm, the compression speed of the algorithm is comparable to that of the SPIHT algorithm, but the image quality (PSNR) under the same compression ratio is 0.2 to 0.2 to higher than that of the SPIHT algorithm. 0.4dB. The algorithm has image compression efficiency comparable to SPIHTwAAC algorithm and JPEG2000 algorithm with relatively low complexity.

Its specific technical solutions are:

An efficient image compression algorithm, including the following steps:

input: image;

Output: Compressed bitstream.

Step 1: Perform wavelet change on the image to generate a set of wavelet coefficients of the pyramid structure;

Step 2: Perform bit-level coding on the image after wavelet change.

Further, the second step is specifically:

First declare some definitions:

LSP: list of important wavelet coefficients;

LIP: list of minor wavelet coefficients;

LIS: list of minor collections;

D(i,j) set: descendants of node (i,j);

L(i, j) set: other descendants of node (i, j) except the first generation descendants;

1. Initialization: output n=log ₂ (max _{(i, j)} | c _{i, j} |); set LSP to an empty list, add the node at the highest level of the pyramid to LIP, and put the D set of descendant nodes into the into the LIS.

2. Classification processing;

3. Refinement processing: For each entry in the LSP (except those entries processed in the last classification), output the value of their nth bit.

4. n=n-1, then go back to step 2.

Further, step 2 is specifically:

2.1 For each node c _i,j in the LIP do:

If| _ci,j |>=2 ⁿ , add _ci,j to the LSP, and then output 1 and the sign bit of _ci,j .

else output 0

end if

end for

2.2 For each set s _i,j in LIS do:

if s _i,j is D set then

if s _i,j has important values,

then outputs 1 and encodes its four first-generation descendants with the D-class set encoding scheme.

If there is no significant first-generation descendant, set _si,j to be a 1-type L set and add it to the LIS set list.

Else s _i,j is a 2-type L set, which is also added to the LIS set list.

end if

Else output 0

end if

If s _i,j is a set of type L then

A class of L-set encoding scheme is used to encode the presence of significant values for its four branches. And move which branches with no significant value to the LIS set list as a new D set.

for those branches with important values;

encode it with a class D set encoding scheme,

If there is no significant first-generation descendant, the branch is a 1-type L set and it is added to the LIS set list.

Else this branch is a 2-type L set, which is also added to the LIS set list.

end if

If s _i,j is a 2-type L set then

The presence of significant values for its four branches is coded with a 2-class L-set coding scheme.

If s _i,j has no significant value, add the 2-type L set s _i,j to the top of the LIS set list.

else then

Add those branches with no significant values to the top of the list of LIS sets as D sets.

For those branches with significant values, they are encoded with the D-class set encoding scheme.

If the first generation descendant of the branch has no significant value, then add the branch to the end of the LIS list as a 1-type L set.

Else adds this branch to the end of the LIS list as a 2-type L set.

end if

end if

end if

end for.

Compared with the prior art, the beneficial effects of the present invention:

The compression speed of the algorithm of the present invention is comparable to that of the SPIHT algorithm, but the image quality (PSNR) under the same compression ratio is 0.2 to 0.4 dB higher than that of the SPIHT. The algorithm has image compression efficiency comparable to SPIHTwAAC algorithm and JPEG2000 algorithm with relatively low complexity. The invention has the characteristics of relatively low complexity and high compression rate, and is suitable for popularization and application.

Detailed ways

The technical solutions of the present invention will be described in further detail below in conjunction with specific embodiments.

An efficient image compression algorithm, including the following steps:

input: image;

Output: Compressed bitstream.

Step 1: Perform wavelet change on the image to generate a set of wavelet coefficients of the pyramid structure;

Step 2: Perform bit-level coding on the image after wavelet change.

Further, the second step is specifically:

First declare some definitions:

LSP: list of important wavelet coefficients;

LIP: list of minor wavelet coefficients;

LIS: list of minor collections;

D(i,j) set: descendants of node (i,j);

L(i, j) set: other descendants of node (i, j) except the first generation descendants;

1. Initialization: output n=log ₂ (max _{(i, j)} | c _{i, j} |); set LSP to an empty list, add the node at the highest level of the pyramid to LIP, and put the D set of descendant nodes into the into the LIS.

2. Classification processing

2.1 For each node c _i,j in the LIP do:

If| _ci,j |>=2 ⁿ , add _ci,j to the LSP, and then output 1 and the sign bit of _ci,j .

else output 0

end if

end for

2.2 For each set s _i,j in LIS do:

if s _i,j is D set then

if s _i,j has important values,

then outputs 1 and encodes its four first-generation descendants with the D-set encoding scheme

If there is no significant first-generation descendant, set _si,j to be a 1-type L set and add it to the LIS set list.

Else s _i,j is a 2-type L set, which is also added to the LIS set list.

end if

Else output 0

end if

If s _i,j is a set of type L then

The presence of significant values in its four branches is encoded with a class-1 L-set encoding scheme. And move which branches with no significant value to the LIS set list as a new D set.

for those branches with important values;

encode it with a D-set encoding scheme,

If there is no significant first-generation descendant, the branch is a 1-type L set and it is added to the LIS set list.

Else this branch is a 2-type L set, which is also added to the LIS set list.

end if

If s _i,j is a 2-type L set then

The presence of significant values for its four branches is coded with a 2-class L-set coding scheme.

If s _i,j has no significant value, add the 2-type L set s _i,j to the top of the LIS set list.

else then

Add those branches with no significant values to the top of the list of LIS sets as D sets.

For those branches with significant values, they are encoded with a D-set encoding scheme.

If the first generation descendant of the branch has no significant value, then add the branch to the end of the LIS list as a 1-type L set.

Else adds this branch to the end of the LIS list as a 2-type L set.

end if

end if

end if

end for.

3. Refinement processing: For each entry in the LSP (except those entries processed in the last classification), output the value of their nth bit.

4. n=n-1, then go back to step 2.

D-set encoding scheme:

Use '0', '1' to indicate whether it is an important descendant, if the four descendants are 0000, the code is 000; if the four descendants are 0001, the code is 001; if the four descendants are 0010, the code is 010; If the four descendants are 0100, the code is 011; if the four descendants are 1000, the code is 100; if the four descendants are 0011, the code is 1010; if the four descendants are 0110, the code is 1011; if the four descendants are 0110, the code is 1011; When the number of descendants is 0101, the code is 1100; if the four descendants are 1001, the code is 1101; if the four descendants are 1010, the code is 11100; if the four descendants are 1100, the code is 11101; if the four descendants are 1100, the code is 11101; When it is 1110, the code is 111100; if the four descendants are 1101, the code is 111101; if the four descendants are 1011, the code is 111110; if the four descendants are 0111, the code is 1111110; if the four descendants are 1111 , the code is 1111111.

Class 1 L-set encoding scheme:

Use '0' and '1' to indicate whether the four branches have important descendants. If the four branches are 0001, the code is 11; if the four branches are 0010, the code is 10; if the four branches are 0100, the code is 011; If the four branches are 1000, the code is 010; if the four branches are 0011, the code is 00111; if the four branches are 0110, the code is 00110; if the four branches are 0101, the code is 00100; if the four branches are 1001 , the code is 00100; if the four branches are 1010, the code is 00010; if the four branches are 1100, the code is 00011; if the four branches are 1110, the code is 000011; if the four branches are 1101, the code is 000010; if If the four branches are 1011, the code is 000001; if the four branches are 0111, the code is 0000001; if the four branches are 1111, the code is 0000000.

2-type L-set encoding scheme:

Use '0' and '1' to indicate whether the four branches have important descendants. If the four branches are 0000, the code is 0; if the four branches are 0001, the code is 1000; if the four branches are 0010, the code is 1001; If the four branches are 0100, the code is 1011; if the four branches are 1000, the code is 1010; if the four branches are 0011, the code is 11000; if the four branches are 0110, the code is 11001; if the four branches are 0101 , the encoding is 11010; if the four branches are 1001, the encoding is 11011; if the four branches are 1010, the encoding is 11101; if the four branches are 1100, the encoding is 111000; if the four branches are 1110, the encoding is 111001; if If the four branches are 1101, the code is 111100; if the four branches are 1011, the code is 111101; if the four branches are 0111, the code is 111110; if the four branches are 1111, the code is 111111.

Table 1 Comparison of image compression quality PSNR between the proposed algorithm and SPIHT algorithm

Table 2. Comparison of running times of images LENA 512X512

We test the proposed algorithm, and the images selected for the test are from a classic image database in academia: USC-SIPI image Database (http://sipi.usc.edu/database/ ) and the Center for Image Processing Research at Rensselaer Polytechnic Institute (CIPR still imageslibrary) (http://www.cipr.rpi.edu/resource/stills/index.html).

It can be seen from Table 1 that the compression quality (PSNR) of the proposed image compression algorithm is better than SPIHT by 0.2dB to 0.4dB under different compression ratios. From Table 2, it can be seen that the proposed image compression algorithm The speed is comparable to that of the SPIHT algorithm.

The above are only preferred specific embodiments of the present invention, and the protection scope of the present invention is not limited thereto. Any person skilled in the art can obviously obtain the simplicity of the technical solution within the technical scope disclosed in the present invention. Variations or equivalent substitutions fall within the protection scope of the present invention.

Claims (1)

1. An efficient image compression algorithm, comprising the steps of:

inputting: an image;

step one, performing wavelet change on an image to generate a wavelet coefficient set with a pyramid structure;

step two, carrying out bit level coding on the image after the wavelet change;

and (3) outputting: compressing a bit stream;

the second step is specifically as follows:

first, some definitions are declared:

LSP: a list of significant wavelet coefficients;

LIP, secondary wavelet coefficient list;

LIS secondary set list;

d (i, j) set: descendants of nodes (i, j);

set of L (i, j): nodes (i, j) other descendants than the first generation descendant;

step 1, initialization: output n is log₂(max_(i,j)|c_i,jI)); setting the LSP as a null list, adding the nodes at the highest level of the pyramid to the LIP, and putting the D set of the nodes with descendants into the LIS;

step 2, classification processing;

step 3, refining treatment: for each table entry which is not classified in the LSP, outputting the value of the nth bit;

step 4, n is n-1, and then the step 2 is returned;

the step 2 specifically comprises the following steps:

step 2.1for each node c in LIP_i,jThe method comprises the following steps:

if | c_i,j|>＝2ⁿC is mixing_i,jAdded to LSP and then output 1 and c_i,jThe sign bit of (a);

otherwise, outputting 0;

step 2.2 Each set s in LIS_i,jThe method comprises the following steps:

if s is_i,jIs the D set

If s is_i,jThere is an important value in (1) for,

outputting 1 and encoding four first generation descendants thereof with a class D set encoding scheme;

if there is no significant first generation descendant, s is determined_i,jSetting a type L set as 1, and adding the type L set into an LIS set list;

otherwise s_i,jIs a type 2L set, and is also added into the LIS set list;

otherwise, outputting 0;

if s is_i,jIs 1 type L set

Coding the condition that the four branches of the L-type set coding scheme have important values by using a type 1L set coding scheme; and moving the branches without important values as a new D set into an LIS set list;

for those branches with significant values;

encoding it with a class D set encoding scheme;

if there are no significant first generation descendants, then the branch is a type 1L set and added to the LIS set list;

otherwise, the branch is a type 2L set and is also added into the LIS set list;

if s is_i,jIs type 2L set

Coding the condition that four branches of the L-type set coding scheme have important values by using a 2-type L set coding scheme;

if s is_i,jWithout significant values, the 2 types L are aggregated s_i,jAdding to the top of the list of LIS collections;

otherwise, adding the branches without important values as D sets to the top of the LIS set list;

for those branches with important values, encoding them with a class D set encoding scheme;

if the first generation descendant of the branch has no significant value, adding the branch to the end of the LIS list with a 1 type L set of identities;

otherwise, the branch is added to the last of the LIS list with the identity of the type 2L set;

wherein, the D set encoding scheme:

with '0', '1' to indicate whether it is an important descendant, if four descendants are 0000, it is encoded as 000; if the four descendants are 0001, encoding 001; if the four descendants are 0010, the code is 010; if the four descendants are 0100, the code is 011; if the four descendants are 1000, then the code is 100; if the four descendants are 0011, the code is 1010; if four descendants are 0110, encode 1011; if the four descendants are 0101, 1100 is encoded; if the four descendants are 1001, encoding 1101; if the four descendants are 1010, the code is 11100; if the four descendants are 1100, the code is 11101, if the four descendants are 1110, the code is 111100; if the four descendants are 1101, encoding 111101; if the four descendants are 1011, encoding 111110; if the four descendants are 0111, 1111110 is encoded; if the four descendants are 1111, the code is 1111111;

class 1L set coding scheme:

indicating with '0' and '1' whether the four branches have significant descendants, if they are 0001, encoding 11; if the four branches are 0010, the code is 10; if the four branches are 0100, the code is 011; if the four branches are 1000, the code is 010; if the four branches are 0011, the code is 00111; if the four branches are 0110, the code is 00110; if the four branches are 0101, the code is 00100; if the four branches are 1001, the code is 00100; if the four branches are 1010, the code is 00010; if the four branches are 1100, the code is 00011; if the four branches are 1110, the code is 000011; if the four branches are 1101, the code is 000010; if the four branches are 1011, the code is 000001; if the four branches are 0111, the code is 0000001; if the four branches are 1111, the code is 0000000;

class 2L set coding scheme:

indicating with '0' and '1' whether the four branches have significant descendants, if the four branches are 0000, encoding 0; if four branches are 0001, the code is 1000; if the four branches are 0010, the code is 1001; if the four branches are 0100, the code is 1011; if the four branches are 1000, the code is 1010; if the four branches are 0011, the code is 11000;

if the four branches are 0110, the code is 11001; if the four branches are 0101, the code is 11010; if the four branches are 1001, the code is 11011; if the four branches are 1010, the code is 11101; if the four branches are 1100, the code is 111000; if the four branches are 1110, the code is 111001; if the four branches are 1101, the code is 111100;

if the four branches are 1011, the code is 111101; if the four branches are 0111, the code is 111110; if the four branches are 1111, the code is 111111.