KR20040014053A - Mobile station and face detection method for mobile station - Google Patents

Abstract

PURPOSE: A mobile communication terminal and a face region extracting method in a mobile communication terminal are provided to efficiently mount a face region extracting engine and a background conversion engine on a DSP chip for the mobile communication terminal, and to share a small memory, thereby minimizing usage of the memory and enabling a real-time processing by using an internal memory of the DSP chip. CONSTITUTION: A memory is divided into an external memory(110) and an internal memory(162). Program codes and data recorded in the external memory(110) are transmitted to a DSP chip(160) through a traffic controller(140) added with a DMA(Direct Memory Access) function. An instruction cache(161) quickly performs programs. Program codes of an encoder, a decoder, a face region extracting engine, a background conversion engine, and an ROI(Region Of Interest)-based bit rate control engine are assigned to the external memory(110), respectively. Infrequent data are assigned to the external memory(110). Frequent data are assigned to the internal memory(162). The program codes of the face region extracting engine and the background conversion engine are assigned to the external memory(110).

Description

A method for extracting a face region in a mobile communication terminal and a mobile communication terminal {Mobile station and face detection method for mobile station}

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a mobile communication terminal, and more particularly, to a mobile communication terminal for image communication capable of performing region of interest (ROI) based bit rate control communication and background switching communication by processing face region extraction in real time.

Today, the mobile communication field has become one of the largest fields representing the information communication field with the peak of IMT2000. Mobile communication services are developing qualitatively based on the higher bandwidth and additional services are also diversifying. Accordingly, the role of the mobile communication terminal is also becoming important, and a miniature mobile communication terminal capable of consuming a small amount of power while providing various functions has emerged.

The main processing functions used in the mobile communication terminal are to encode the data to be transmitted or to decode the received data. Unlike terminals that only transmit and receive audio data, mobile terminals for video communication, which have recently emerged, need to encode / decode video images. Encoding / decoding video images requires more throughput than encoding / decoding video data. Therefore, in order to perform real-time processing faster, it is necessary to use dedicated hardware, and a DSP (Digital Signal Processing) chip is used. Such a DSP chip is a dedicated chip that can quickly process a signal processing that is simple but has many repetitive operations such as a discrete cosine transform (DCT) transform used in encoding / decoding. Recently, most mobile communication terminals are equipped with such a DSP chip. The biggest feature of the DSP chip for a mobile communication terminal is that a large number of processing can be performed using a low power supply.

Representative products of DSP chips in such mobile communication terminals include the TMS320C52x series and the TMS320C55x series from Texas Instruments. These DSP chips feature low power consumption while maintaining performance from 150 to 200 MHz. In addition, these DSP chips have limited but internal memory, which is generally advantageous over using external memory because the memory can be accessed once per chip cycle. In the case of a program, a specific DSP chip is equipped with a program cache, so that even if the program code is in an external memory, the program cache can be used to quickly operate. As a result, complex algorithms can be processed in real time only when the DSP chip features such as limited internal memory and program cache are utilized.

In addition, the recent trend of the DSP chip for a mobile communication terminal is changing from a specific hardware-dependent design to a general-purpose design. TI's C54x and C55x are also DSP chips for mobile handsets that can perform these functions. That is, it can be designed to adapt well to the characteristics of mobile communication terminals developed by various companies. Due to this general characteristic, the DSP chip can be utilized more by performing various processing using the DSP chip.

On the other hand, in the field of video coding research, there is a research on a technology for extracting a face region and coding a face region with higher image quality and other background regions with a lower image quality to transmit and receive a high quality image with a small amount of data. A lot is being done.

Since the face area is a region of interest (ROI) in video communication, if only the face area is high quality, the user may feel less inconvenience in image quality. At this time, it is important to extract the face region engine for the face region extraction. The face region extraction method has been studied for a long time. There are many ways of extracting the face area, including how to use face color information, how to use feature information such as eyes, nose and mouth, and template matching methods to construct and compare face templates in advance. have. Among them, the most commonly used method is the use of face color information, which is relatively fast and shows good performance.

However, most of the face region extraction techniques introduced above focus only on accurately extracting a face region in a personal computer environment, and when applied to a mobile communication terminal environment, it takes a lot of processing time to take several seconds to severely. Therefore, in a mobile communication terminal which requires a real-time processing while using limited hardware, a face region extraction algorithm should be developed to have a good structure to use a DSP chip.

At this time, there are two major considerations when performing the face region extraction algorithm in the DSP chip. First, in terms of CPU utilization, you should use as few CPUs as possible. To do this, the algorithm itself must be very simple, and the algorithm must consist of simple iterative processing that the DSP chip can process quickly. The next consideration is the use of less memory. Most DSP chips use internal memory for fast processing, so the amount of data used should be small. This requires an optimized memory design that can analyze the data memory required in the processing time sequence and share the same memory. Similarly, the program code area occupies the memory. It is also recommended to use internal memory. However, if there is a program cache like C55x, the program code can be loaded quickly in external memory.

The present invention provides a mobile communication terminal capable of processing face region extraction in real time by developing a face region extraction algorithm in consideration of characteristics of a DSP chip and designing a memory allocation to be optimized for a face region extraction algorithm. The purpose is.

In addition, the present invention is equipped with a face region extraction engine, a bit rate control engine, and a background switching engine in addition to an encoder / decoder basically performed by using a DSP chip in a mobile communication terminal equipped with a DSP chip, so that even in a low network environment, Another purpose is to provide a mobile communication terminal capable of maintaining and providing services such as background switching.

1 is a block diagram schematically showing the configuration of a mobile communication terminal according to the present invention;

2 is a view for explaining a process of extracting a face region by a face region extraction method in a mobile communication terminal according to the present invention.

3 is a diagram illustrating a memory usage state in each process of extracting a face region by a face region extraction method in a mobile communication terminal according to the present invention;

4 is a diagram illustrating an example of a screen on which background switching communication is performed in a mobile communication terminal according to the present invention;

5 is a view for explaining a process of performing a background switching communication in a mobile communication terminal according to the present invention.

6 conceptually illustrates memory area allocation of a mobile communication terminal according to the present invention;

<Explanation of symbols for the main parts of the drawings>

110 ... external memory 120 ... camera

130 ... LCD 140 ... Traffic control unit

150 ... frame buffer 160 ... DSP

Instruction cache 162 Internal memory

In order to achieve the above object, a mobile communication terminal according to the present invention,

A first memory area allocated as a storage area for data for 'color image' processing and data for 'Dilation image' processing;

A second memory area allocated as a storage area for data for 'index image' processing and data for 'Erosion image' processing;

A third memory area allocated as a storage area for data for 'gray image' processing;

A fourth memory area allocated as a storage area of a global variable necessary for the operation in the face region extraction process;

A fifth memory area allocated as a storage area of a local variable necessary for an operation in a face area extraction process; And

A digital signal processing (DSP) chip having a face region extraction engine for extracting a face region from an input image using data stored in the first to fifth memory regions; Its features are to include.

Here, the 'index image' represents an image composed of color index values of pixels satisfying the skin color condition in the input image, and the 'skin image' is a skin color by grouping colors around the main skin color from the 'index image'. The image represents only the pixels belonging to the group. The gray image represents an image obtained by converting the input image to gray. The 'Erosion image' represents an image from which flesh color pixels generated by noise are removed from the 'color image'. The 'Dilation image' represents an image configured by filling small holes in the flesh color region generated by noise in the 'Erosion image'.

Further, according to the present invention, the code part of the face region extraction engine is located in an external memory of the DSP chip, and the first to fifth memory areas are located in an internal memory of the DSP chip.

According to the present invention, all of the first to third memory areas are 88 * 72 Word size, the fourth memory area is 890 Word size, the fifth memory area is 100 Word size Has its features.

According to the present invention, the apparatus further comprises a sixth memory area storing a background image to be used as a background of a screen on which an image is displayed, and the DSP chip further includes a background switching engine for selectively synthesizing the input image with the background image. There is a characteristic in that.

According to the present invention, the sixth memory area and the code part of the background switching engine are characterized in that they are located in an external memory of the DSP chip.

In addition, according to the present invention, a seventh memory area for storing the original image to be used as an input image is further provided in the external memory of the DSP chip, and the original image and the background image in macroblock units to synthesize the original image and the background image This feature is characterized in that the synthesis in the internal memory of the DSP chip.

In addition, according to the present invention, the DSP chip further comprises a ROI-based bit rate control engine for controlling the bit rate relatively different depending on the ROI so as to be set to a relatively high picture quality relative to the non-ROI area for the main ROI. It has that feature.

In addition, in order to achieve the above object, the method for extracting a face region in a mobile communication terminal according to the present invention,

Investigating whether all the pixels of the input image satisfy the skin color condition;

Constructing an 'index image' composed of index values by obtaining an index of color values of each pixel for pixels satisfying the flesh color condition;

Comprising the most common color in the 'index image' grouping similar colors into a group to 'On' only the pixels corresponding to the color group, and the other 'Off' to configure a 'color image' Wow;

Constructing an 'Erosion image' by performing an 'Erosion morphology' on the 'color image';

Constructing a 'dilation image' by performing a 'dilation morphology' on the 'Erosion image';

A connecting element constituting step of connecting the on-pixels connected to each other in the 'Dilation image' as a single mass and displaying the connected area;

Designating as a face candidate region an element satisfying a predetermined condition among the connection elements configured in the connection element configuring step;

An elliptic matching step of examining whether the designated face candidate region has an ellipse shape;

Converting the first input image into a gray image to construct a 'gray image';

Irradiating an eye and a mouth area around the face candidate area in the 'gray image'; And

And a face region determination step of determining a face region using the elliptic matching result and the eye and mouth region findings.

According to the present invention, the same memory area is allocated to the data for processing the 'color image' and the 'Dilation image' and shared according to time, and the data for the 'index image' and 'Erosion image' processing are used. The same memory area may be allocated and shared according to time, and the data for the 'gray image' processing may be allocated by using a separate memory area.

In addition, in order to achieve the above object, the mobile communication terminal according to the present invention includes a DSP including a face region extraction engine for extracting a face region and a background switching engine for selectively synthesizing an input image and a background image. A chip; A code section of the face region extraction engine mounted in an external memory of the DSP chip; A global variable unit mounted in an internal memory of the DSP chip and used by the face region extraction engine for face region extraction; A local variable unit mounted in an internal memory of the DSP chip and used by the face region extraction engine for face region extraction; A code unit of the background switching engine mounted in an external memory of the DSP chip; A global variable unit used by the background switching engine mounted in an internal memory of the DSP chip; And a local variable unit used in the background switching engine mounted in the internal memory of the DSP chip.

Based on the extracted face region information, there is a feature in that a background conversion is performed by selectively synthesizing the input image and the background image.

In addition, another example of a mobile communication terminal according to the present invention for achieving the above object is a face region extraction engine for extracting a face region, and a bit rate according to whether the region of interest (ROI) or not; A DSP chip equipped with a ROI-based bit rate control engine for differentially controlling the signal; A code section of the face region extraction engine mounted in an external memory of the DSP chip; A global variable portion used by the face region extraction engine mounted in an internal memory of the DSP chip; A local variable unit used by the face region extraction engine mounted in an internal memory of the DSP chip; And a code unit of the ROI-based bit rate control engine mounted in an external memory of the DSP chip. It is configured to include,

Based on the extracted face region information, the feature is that the bit rate is controlled relatively differently according to whether or not the face region is set so that the face region is set to a higher image quality than the non-face region.

According to the present invention, there is an advantage that the communication can be performed by extracting the face region in real time and controlling the bit rate or switching the background based on the ROI.

In addition, according to the present invention, a face region extraction engine and a background switching engine are effectively mounted on a DSP chip for a mobile communication terminal having a characteristic of low memory and limited power, and in particular, it is designed to share a small amount of memory to minimize memory usage. In addition, the main data has an advantage of enabling real-time processing by using the internal memory of the DSP chip.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

The present invention proposes a DSP-based mobile communication terminal equipped with a face region extraction engine, an ROI-based bit rate control engine, a background switching engine, an encoder, and a decoder. In the embodiment of the present invention, the TMS320C55x of TI Corp. is used as the DSP chip, and the entire frame structure will be described based on the case of adopting the OMAP1510 structure in which two CPUs, C55x and ARM925 ^TM CPU, are mounted. However, the technical idea of the present invention is not limited thereto and should be interpreted based on the contents described below.

In the case of the OMAP1510, the internal memory of the C55x has 80 kWords (160 KB), 1.5 Mbits, on-chip memory, which will be used as buffering to display frames on the LCD of the mobile communication terminal. Is equipped. It is also equipped with a 24KB program cache and the DSP chip itself is capable of 170MHz to 200MHz. All data processing of C55x is processed in Word unit. Unless otherwise processed, the smallest data unit is basically 1 Word (2Byte). The mobile communication terminal to which the present invention is applied further includes an external memory in the OMAP1510. The external memory contains the data or program code that the DSP chip cannot use in the internal memory and the data and program code for the ARM CPU.

<Design of mobile communication terminal>

The configuration of a mobile communication terminal according to the present invention will be described with reference to FIG. 1. 1 is a diagram illustrating an example of a configuration of a mobile communication terminal capable of ROI-based bit rate control and background switching communication according to the present invention.

Referring to FIG. 1, when divided based on the DSP chip 160, the memory is largely divided into an external memory 110 and an internal memory 162. The program code and data recorded at 110 may be transferred into the DSP chip 160 through the traffic controller 140 to which a direct memory access (DMA) function is added. In addition, the camera 120 is connected to the mobile communication terminal for image communication, and a dedicated buffer 150 is mounted to decode the received image and display the received image on the liquid crystal display 130. In addition, an instruction cache (I-Cache) 161 is mounted in the DSP chip 160 to execute a program more quickly.

The external memory 110 is allocated program codes of an encoder, a decoder, a face region extraction engine, a background switching engine, and an ROI-based bit rate control engine, respectively. In this case, the program code may use a program cache of C55x, so that even if the program code is allocated to the external memory 110, fast processing is possible. In addition, the data area of the encoder / decoder is allocated to the external memory 110 (Enc./Dec.data1 area) for data that is not frequently used, and to the internal memory 162 for data that is frequently used. (Enc./Dec.data2 area) Allocates. However, in the present invention, the encoder and the decoder do not limit the mounting method because all mobile communication terminals must be installed in common.

In the face region extraction engine and the background switching engine, program codes are allocated to the external memory 110 and data to be used are allocated to the internal memory 162. In this case, the data used by the face region extraction engine and the background switching engine are designed to share the same memory location by analyzing life-cycles in order of processing time. The detailed memory design details will be described later.

DSP Design of Face Region Extraction Engine

Before the face region extraction engine is mounted on the DSP, we will look at the face region extraction algorithm. The sequence diagram for the face region extraction algorithm is shown in FIG. 2, and for the DSP design, the sequence diagram of all algorithms is described based on the data flow generated or input. The face region extraction algorithm is largely composed of a color-based face candidate region extraction module and a feature-based face region extraction module.

The first stage is the input stage of the image. Frames through the camera are stored in an external frame buffer, and are processed in macroblock units in order in the internal memory. In other words, the entire image of 176 * 144 * 1.5 Words occupies about 50% of the limited memory when it is processed in the internal memory at once. Since the original frame is not processed repeatedly and read only once at first, it is not used afterwards, so it can be loaded into internal memory in macroblock units. In the macroblock image loaded in the memory, each pixel is checked whether the color satisfies the skin color range, and if the color is the skin color range, the color block is quantized, converted into a color index, and marked on the corresponding part of the 'index image'.

Conversion of the color index means that a number of colors represented in the YcbCr color space are grouped in the color space to display colors of the same group as one index. In the present invention, all the colors in the flesh color range are represented by 72 indexes. The 'index image' consists of 88 * 72, the original size of 176 * 144, reduced by half for horizontal and vertical. This can be reduced in size by inspecting pixels of the original image, which are raised in units of macro blocks, by crossing each column horizontally and vertically.

In addition to constructing an 'index image', each pixel to be examined is converted to gray to form a 'gray image'. The gray image is also 88 * 72. Putting all the macroblocks into the internal memory in turn, this operation consists of 'index memory' and 'gray memory'.

Color grouping is performed again using the index information stored in the 'index memory'. Here, the color grouping means that similar colors are grouped into one group based on the color most distributed among the flesh color pixels extracted from the current image. The method of grouping similar colors is performed by constructing a color histogram of 72 color indexes as bins, and then grouping indexes representing similar colors into one centered around the color represented by the bin having the largest value. After the grouping process, only the index belonging to the group is 'On' and the rest is composed of 'color image'.

The 'color image' is composed of pixels, and includes a lot of noise. Among the noises are noises that appear 'On' by one or two pixels on the background and noises represented by small holes in the face area. First, the 'Erosion morphology' is used to construct an 'Erosion image' with noise removed from the background. If the 'Dilation morphology' is used again for the 'Erosion Image', the 'Dilation Image' is filled with the small holes generated by the noise.

After dividing the 'Dilation Image' into a 4 * 4 grid again, if only one grid has a certain number of 'On' pixels above a certain threshold, 'On' and if it is 'Off' the other 22'18 The grid image consists of a grid. Then, when the grids 'On' in the 'grid image' are adjacent to each other, a connection component is displayed which displays them as one connected component.

That is, they define adjacent chunks by tying adjacent grids together. These bundled components make up each subregion in the image, and one of these subregions is more likely to be a face. Therefore, if a component having a predetermined size or more among the connected components is designated as the face candidate region, the feature-based face region extraction module will check and readjust it to extract the face region. In this case, when the image is not complicated, even if the feature-based face region extraction module is not present, the connection component itself extracted based on the color may be extracted as the face region.

The feature-based face region extraction module basically uses 'Dilation Image' and 'Gray Image', which are previously skinned images without noise. Elliptic matching is performed to check whether the distribution of flesh color is close to an ellipse using a 'dilation image'. In general, the face is elliptical, and thus the probability of being a face is increased when it is distributed close to the ellipse. Eyes and mouths are examined inside the ellipse using the 'gray image'. Eyes and mouths have a dark, horizontal line pattern. If the conditions of the eyes, mouth, and ellipse are satisfied, the elliptical region is finally defined as the facial region.

In order to mount the above-described face region extraction algorithm on the DSP chip, the program code is first allocated to an external memory as described above. The processed data is 88 * 72 Word of 'Index Image', 'Gray Image', 'Color Image', 'Erosion Image', and 'Dilation Image', and other data used for processing such as 'Grid Image' They use approximately 890 words. These are declared as global variables. Five images using 88 * 72 words occupy the largest area, and they are designed to effectively share and use memory. To this end, the life cycle of each process is defined as shown in FIG. 3.

Referring to FIG. 3, when the memory area used by a variable that is no longer used is designed to be used by another variable, it can be seen that only three memory areas of 88 * 72 Word size can effectively use the memory. Let's call these three regions 'Image1', 'Image2', and 'Image3' respectively.

First, 'Image1' is used as an area for one macro block image loaded in the internal memory when uploading from the original image in the external memory to the internal memory in macro block units. After that, it is used as 'color image' and 'Dilation image' area after applying 'Dilation morphology'. And, 'Image2' is used as an area for 'index image' and 'Erosion image' after 'Erosion morphology' is applied.After completion of color-based face candidate area extraction module, global used by feature-based face area extraction module Used for some of the variables. Also, 'Image3' is used for 'gray image'. 'Gray images' are first created and used for eye and mouth areas later in the face area extraction, so they have the longest life cycle.

In this way, by sharing three 88 * 72 Word memory areas and using 890 Word's internal memory as other global variables, all global variables can be allocated to the internal memory. A small amount of local variables used temporarily are allocated and used as an area called a stack in the case of C55x. In the present invention, a stack for local variables of 100K Word is used. In this way, the total memory size is about 20K words, and only 25% of the DSP's total internal memory is available.

In the above-described method, all memory allocations required for DSP loading of the face region extraction engine were designed. The face region information extracted by the face region extraction engine may be used to control bit rate or perform background switching communication based on ROI. ROI-based bit rate control refers to a control technique that maintains a high quality image for a face region, which is a major area of interest, and maintains a high quality face region for communication. ROI-based bit rate control may be performed using face region information in its own bit rate control portion of the encoder.

And background switching communication means the function which communicates by replacing the area | region other than a face with another still image, as shown in FIG. Background switching communication enables various useful and interesting communication, and at the same time, the background is a still image, so it can be compressed and communicated in a very small amount. The background conversion engine uses the extracted face region information to synthesize a background portion as a background portion selected by a user and a face region portion as a portion input from a camera and input the same to an encoder. Since the encoder receives and processes input in units of macro blocks, the background conversion engine also synthesizes images in units of macro blocks. That is, the background image is in the external memory, and when the user selects the background, the macro block of the background is uploaded to the internal memory, and if the macro block includes the face area, the macro block from the camera image is also uploaded to the internal memory. Next, the two images are synthesized.

In this case, two macro block sizes (16 * 16 * 1.5) that are uploaded to the internal memory may be shared by using one of the above-described 'Image1', 'Image2', and 'Image3'. The synthesized macro block size synthesized image is encoded into the input of the encoder. As one method of performing such encoding, encoding may be performed using an H.263 video encoder. The processing on the DSP chip of the background switching engine described above is shown in FIG. 5.

The configuration of ROI-based bit rate control and background switching mobile communication terminal using the TI320C55x DSP chip described so far is shown in FIG. 6. The external memory is allocated program code of each engine such as face region extraction engine, bit rate control engine, background switching engine, encoder, decoder, etc., and internal memory is divided into three 88s for global variables used by face region extraction engine and background switching engine. * 72 There is a global variable space ('Image1', 'Image2', 'Image3') of Word size. A 890 Word global variable for other global variables also exists in internal memory, and a 100 word stack for local variables is allocated in internal memory. The background and camera input frames are then allocated to an external memory buffer. In C55x, all global variables have their initial values due to memory management. Initial value information is the same size as global variables and is allocated to another memory named '.init'. Here, '.init' is allocated to external memory because the initial value needs to be read only once when executing the program.

As described above, the mobile communication terminal according to the present invention has an advantage of performing communication by extracting a face region in real time and controlling a bit rate or switching a background based on ROI.

In addition, the mobile communication terminal according to the present invention is particularly designed to share a small amount of memory by effectively mounting a face region extraction engine and a background switching engine in a DSP chip for a mobile communication terminal having characteristics of low memory and limited power. This minimizes the use of memory and enables the real-time processing by using the internal memory of the DSP chip for main data.

Claims (11)

A first memory area allocated as a storage area for data for 'color image' processing and data for 'Dilation image' processing; A second memory area allocated as a storage area for data for 'index image' processing and data for 'Erosion image' processing; A third memory area allocated as a storage area for data for 'gray image' processing; A fourth memory area allocated as a storage area of a global variable necessary for the operation in the face region extraction process; A fifth memory area allocated as a storage area of a local variable necessary for an operation in a face area extraction process; And A digital signal processing (DSP) chip having a face region extraction engine for extracting a face region from an input image using data stored in the first to fifth memory regions; Mobile communication terminal comprising a. Here, the 'index image' represents an image composed of color index values of pixels satisfying the skin color condition in the input image, and the 'skin image' is a skin color by grouping colors around the main skin color from the 'index image'. The image represents only the pixels belonging to the group. The gray image represents an image obtained by converting the input image to gray, and the 'Erosion image' represents an image from which the flesh color pixels generated by noise are removed from the 'color image'. The 'Dilation image' represents an image configured by filling small holes in the flesh color region generated by noise in the 'Erosion image'. The method of claim 1, The code unit of the face region extraction engine is located in an external memory of the DSP chip, and the first to fifth memory areas are located in an internal memory of the DSP chip. The method of claim 1, The first to third memory areas are all 88 * 72 words, the fourth memory area is 890 Word, and the fifth memory area is 100 Word. . The method of claim 1, And a sixth memory area storing a background image to be used as a background of a screen on which an image is displayed, wherein the DSP chip further includes a background switching engine for selectively synthesizing an input image with a background image. Communication terminal. The method of claim 4, wherein And a code part of the sixth memory area and the background switching engine is located in an external memory of the DSP chip. The method of claim 4, wherein A seventh memory area for storing an original image used as an input image is further included in an external memory of the DSP chip, and in order to synthesize the original image and the background image, the original image and the background image in macroblock units are included in the DSP chip. A mobile communication terminal, characterized in that synthesized in the memory. The method of claim 1, The DSP chip further comprises a ROI-based bit rate control engine for controlling the bit rate relatively different according to the ROI so as to be set to a relatively high picture quality relative to the non-ROI area for the main area of interest (ROI). Mobile terminal. Investigating whether all the pixels of the input image satisfy the skin color condition; Constructing an 'index image' composed of index values by obtaining an index of color values of each pixel for pixels satisfying the flesh color condition; Comprising the most common color in the 'index image' grouping similar colors into a group to 'On' only the pixels corresponding to the color group, and the other 'Off' to configure a 'color image' Wow; Constructing an 'Erosion image' by performing an 'Erosion morphology' on the 'color image'; Constructing a 'dilation image' by performing a 'dilation morphology' on the 'Erosion image'; A connecting element constituting step of connecting the on-pixels connected to each other in the 'Dilation image' as a single mass and displaying the connected area; Designating as a face candidate region an element satisfying a predetermined condition among the connection elements configured in the connection element configuring step; An elliptic matching step of examining whether the designated face candidate region has an ellipse shape; Converting the first input image into a gray image to construct a 'gray image'; Irradiating an eye and a mouth area around the face candidate area in the 'gray image'; And And a face region determination step of determining a face region by using the elliptic matching result and the eye and mouth region findings. The method of claim 8, The same memory area is allocated to the data for processing the 'color image' and the 'Dilation image' and shared according to time, and the same memory area is used for the data for the 'index image' and 'Erosion image' processing. The method for extracting a face region in a mobile communication terminal, wherein the data is allocated and shared according to time, and the data for processing the 'gray image' is allocated by using a separate memory area. A DSP chip having a face region extraction engine for extracting a face region, and a background switching engine for selectively synthesizing an input image and a background image; A code section of the face region extraction engine mounted in an external memory of the DSP chip; A global variable unit mounted in an internal memory of the DSP chip and used by the face region extraction engine for face region extraction; A local variable unit mounted in an internal memory of the DSP chip and used by the face region extraction engine for face region extraction; A code unit of the background switching engine mounted in an external memory of the DSP chip; A global variable unit used by the background switching engine mounted in an internal memory of the DSP chip; And a local variable unit used in the background switching engine mounted in the internal memory of the DSP chip. And based on the extracted face region information, selectively converts the input image and the background image to perform background switching. A DSP chip including a face region extraction engine for extracting a face region and a ROI-based bit rate control engine for controlling bit rates relatively differently according to a region of interest (ROI); A code section of the face region extraction engine mounted in an external memory of the DSP chip; A global variable portion used by the face region extraction engine mounted in an internal memory of the DSP chip; A local variable unit used by the face region extraction engine mounted in an internal memory of the DSP chip; And A code unit of the ROI based bit rate control engine, mounted in an external memory of the DSP chip; It is configured to include, And based on the extracted face region information, controlling the bit rate relatively differently according to whether or not the face region is set so that the face region is set to a higher quality than the non-face region.