CN108154539A - A kind of colorspace data converting algorithm based on Opengl ES - Google Patents

Abstract

The present invention relates to a kind of colorspace data converting algorithms based on Opengl ES, include the following steps：Tinter shader is write using Opengl shading languages GLSL so that a tinter shader is run on parallel on multiple cores of GPU, improves the speed of colorspace data conversion；It needs to create two kinds of tinters in Opengl ES：Vertex shader vertex shaders and fragment shader fragment shaders.The present invention handles data using GPU, and completing YUV colorspace datas based on Opengl ES is converted into RGBA color spatial data, can increase substantially transformation efficiency, support 1080P（1920x1080）The requirement per second of 30 frame picture of video.

Description

A kind of colorspace data converting algorithm based on Opengl ES

Technical field

The present invention relates to colorspace data transformation technology field, specifically a kind of color based on Opengl ES is empty Between data converting algorithm.

Background technology

Existing mobile hand-held device, in addition to central processing unit（CPU）Except performance is continuously improved, also match accordingly mostly For the graphics processor of profession（GPU）.

Graphics processor（GPU）And central processing unit（CPU）It compares：CPU need very strong versatility handle it is various not Same data type, while have logic judgment again, the processing that a large amount of branch redirects and interrupts can be therefore introduced, these all cause The internal structure complex of CPU.And GPU is faced be then type high unity, the large-scale data mutually without dependence and The pure computing environment being interrupted is not needed to.

It is suitble to the program of operation on GPU：

（1）The program of computation-intensive.So-called computation-intensive（Compute-intensive）Program, be exactly its major part For run time flower on Register operations, the speed of register and the speed of processor are suitable, several from register read-write data It is not delayed.

（2）It is easy to parallel program.GPU is a kind of SIMD frameworks in fact（Single Instruction Multiple Data）, it has hundreds and thousands of a cores, each core can do same thing in the same time.

When being converted into RGBA color spatial data with CPU completion YUV colorspace datas, because of CPU register numbers It is not very much, it is difficult to realize highly-parallel, algorithm（As C++ is realized）Performance is low, it is impossible to meet high definition（Such as 1080P）The reality of video When show.

Invention content

For defect in the prior art, the purpose of the present invention is to provide a kind of colors based on Opengl ES Spatial data converting algorithm handles data using GPU, and completing YUV colorspace datas based on Opengl ES is converted into RGBA face Colour space data can increase substantially transformation efficiency, support 1080P（1920x1080）The requirement per second of 30 frame picture of video.

To achieve the above objectives, the technical solution adopted by the present invention is that：

A kind of colorspace data converting algorithm based on Opengl ES, which is characterized in that include the following steps：

Tinter shader is write using Opengl shading languages GLSL so that a tinter shader runs on GPU parallel Multiple cores on, improve colorspace data conversion speed；

It needs to create two kinds of tinters in Opengl ES：Vertex shader vertex shaders and fragment shader fragment shaders。

Based on the above technical solution, the colorspace data conversion refers to YUV colorspace datas and is converted into RGBA color spatial data.

Based on the above technical solution, the data that the YUV colorspace datas are captured from camera,

Specific form is NV21 forms.

Based on the above technical solution, it is empty to complete YUV colors using following conversion formula by the tinter shader Between data be converted into RGBA color spatial data：

R = Y + 1.13983 *（V - 128）

G = Y - 0.39465 *（V - 128）- 0.58060 * （V - 128）

B = Y + 2.03211 * （U -128）.

Based on the above technical solution, the fragment shader, image sampling is used using 256 as maximum It is worth normalized floating number,

YUV conversions RGB uses the floating number between 0.0 to 1.0 to be represented with matrix.

Based on the above technical solution, the fragment shader,

The principal function interface main called automatically by system is provided,

Texture sample coordinates vTextureCoord is set,

It sets and precision is given tacit consent in vertex shader as precision highp float,

Y texture cell texture are set,

VU texture cell texture2 are set,

Texture texture cells are uploaded to using the grayscale format LUMINANCE of Opengl ES to single pass Y images,

Transparency form LUMINANCE_ALPHA is added to upload to using the gray scale of Opengl ES twin-channel VU images Texture2 texture cells,

By texture sample coordinates vTextureCoord, texture texture cells and texture2 texture cells are adopted respectively Sample obtains two 4 dimensions color vector vec41 and vec42,

4 dimension color vector vec41, the r dimension vec41.r is y pixels,

4 dimension color vector vec42, the r and a dimension vec42.ra is vu pixels,

One y pixel and a vu pixel are formed into three-dimensional color vector v ec3 (vec41.r, vec42.ra),

According to conversion formula, three-dimensional color vector v ec3 is multiplied by transformed matrix mat3 and just obtains a rgb pixels output display.

Colorspace data converting algorithm of the present invention based on Opengl ES handles data using GPU, is based on Opengl ES complete YUV colorspace datas and are converted into RGBA color spatial data, can increase substantially transformation efficiency, support 1080P（1920x1080）The requirement per second of 30 frame picture of video.With the algorithm method of wide high progress dual ring calculation to image of CPU It compares, substantially increases efficiency of algorithm.

Specific embodiment

Colorspace data converting algorithm of the present invention based on Opengl ES, includes the following steps：

Using Opengl shading languages GLSL（Opengl Shader Language）Write tinter shader so that one Color device shader is run on parallel on multiple cores of GPU, improves the speed of colorspace data conversion.

Opengl ES（Opengl for Embedded Systems）It is the subset of Opengl 3-D graphic API, for The embedded devices such as mobile phone, PDA and game host and design.The API defines popularization by Khronos groups, and Khronos is one Figure software and hardware employer's organization, the association are primarily upon the open standard in terms of figure and multimedia.

Tinter（shader）It is that small, the program based on C language is write with shading language.It is needed in Opengl ES Create two kinds of tinters：Vertex shader（vertex shaders）And fragment shader（fragment shaders）.This two Kind tinter is the two halves of a complete routine, it is impossible to only create any of which one；Want to create a complete coloring journey Sequence is both necessarily present.Vertex shader defines how the geometric figure in 2D 3D scenes is handled.One Vertex refers to a point in 2D or 3d space.In image procossing, there are 4 vertex：Each vertex representative image One angle.Vertex shader sets the position on vertex, and parameter as position and texture coordinate is sent to fragment shading Device.It is calculated in each of object or picture pixel using fragment shader, finally calculates each pixel most Whole color.

Based on the above technical solution, the colorspace data conversion refers to YUV colorspace datas and is converted into RGBA color spatial data.

" Y " represents brightness in YUV（Luminance or Luma）, that is, grayscale value；And what " U " and " V " was represented is then Coloration（Chrominance or Chroma）, effect is description colors of image and saturation degree, for the color of specified pixel.I.e.：It is bright Spend signal Y, colour difference signal B-Y（That is U）, R-Y（That is V）.

RGBA represents Red（It is red）、Green（Green）、Blue（Blue）It is that is, transparent with the color space of Alpha Degree/opacity.

Based on the above technical solution, the data that the YUV colorspace datas are captured from camera.

Its specific form can be NV21 forms.NV21 is before V after U.Also referred to as YUV420 frame informations or cry The quantity ratio of tri- components of YCbCr_420_SP, Y, U, V is 4:1:1, that is to say, that every four pixels share a pair of of UV.

The camera is mobile phone camera.

Based on the above technical solution, it is empty to complete YUV colors using following conversion formula by the tinter shader Between data be converted into RGBA color spatial data：

R = Y + 1.13983 *（V - 128）

G = Y - 0.39465 *（V - 128）- 0.58060 * （V - 128）

B = Y + 2.03211 * （U -128）.

Based on the above technical solution, the fragment shader, also known as piece member tinter（Fragment Shader）, image sampling is used using 256 as the normalized floating number of maximum value,

YUV conversions RGB uses the floating number between 0.0 to 1.0 to represent as follows with matrix：

= * 。

I.e.：Using the floating-point matrix of 3*3, i.e. transformed matrix,

mat3（1, -0.00093,1.401687,

1, -0.3437, -0.71417,

1,1.77216,0.00099）.

Based on the above technical solution, the fragment shader,

The principal function interface main called automatically by system is provided,

Texture sample coordinates vTextureCoord is set,

It sets and precision is given tacit consent in vertex shader as precision highp float,

Y texture cell texture are set,

VU texture cell texture2 are set,

Texture texture cells are uploaded to using the grayscale format LUMINANCE of Opengl ES to single pass Y images,

Transparency form LUMINANCE_ALPHA is added to upload to using the gray scale of Opengl ES twin-channel VU images Texture2 texture cells,

By texture sample coordinates vTextureCoord, texture texture cells and texture2 texture cells are adopted respectively Sample obtains two 4 dimensions color vector vec41 and vec42,

4 dimension color vector vec41, the r dimension vec41.r is y pixels,

4 dimension color vector vec42, the r and a dimension vec42.ra is vu pixels,

One y pixel and a vu pixel are formed into three-dimensional color vector v ec3 (vec41.r, vec42.ra),

According to conversion formula, three-dimensional color vector v ec3 is multiplied by transformed matrix mat3 and just obtains a rgb pixels output display.

The content not being described in detail in this specification belongs to the prior art well known to professional and technical personnel in the field.

Claims (6)

1. a kind of colorspace data converting algorithm based on Opengl ES, which is characterized in that include the following steps：

Tinter shader is write using Opengl shading languages GLSL so that a tinter shader runs on GPU parallel Multiple cores on, improve colorspace data conversion speed；

It needs to create two kinds of tinters in Opengl ES：Vertex shader vertex shaders and fragment shader fragment shaders。

2. the colorspace data converting algorithm based on Opengl ES as described in claim 1, it is characterised in that：The face The conversion of colour space data refers to YUV colorspace datas and is converted into RGBA color spatial data.

3. the colorspace data converting algorithm based on Opengl ES as claimed in claim 2, it is characterised in that：The YUV The data that colorspace data is captured from camera,

Specific form is NV21 forms.

4. the colorspace data converting algorithm based on Opengl ES as claimed in claim 3, it is characterised in that：Described Color device shader completes YUV colorspace datas using following conversion formula and is converted into RGBA color spatial data：

R = Y + 1.13983 *（V - 128）

G = Y - 0.39465 *（V - 128）- 0.58060 * （V - 128）

B = Y + 2.03211 * （U -128）.

5. the colorspace data converting algorithm based on Opengl ES as claimed in claim 3, it is characterised in that：Described Section tinter uses image sampling using 256 as the normalized floating number of maximum value,

YUV conversions RGB uses the floating number between 0.0 to 1.0 to be represented with matrix.

6. the colorspace data converting algorithm based on Opengl ES as claimed in claim 5, it is characterised in that：Described Section tinter,

The principal function interface main called automatically by system is provided,

Texture sample coordinates vTextureCoord is set,

It sets and precision is given tacit consent in vertex shader as precision highp float,

Y texture cell texture are set,

VU texture cell texture2 are set,

Texture texture cells are uploaded to using the grayscale format LUMINANCE of Opengl ES to single pass Y images,

Transparency form LUMINANCE_ALPHA is added to upload to using the gray scale of Opengl ES twin-channel VU images Texture2 texture cells,

By texture sample coordinates vTextureCoord, texture texture cells and texture2 texture cells are adopted respectively Sample obtains two 4 dimensions color vector vec41 and vec42,

4 dimension color vector vec41, the r dimension vec41.r is y pixels,

4 dimension color vector vec42, the r and a dimension vec42.ra is vu pixels,

One y pixel and a vu pixel are formed into three-dimensional color vector v ec3 (vec41.r, vec42.ra),

According to conversion formula, three-dimensional color vector v ec3 is multiplied by transformed matrix mat3 and just obtains a rgb pixels output display.