Analysis of the technical advantages of Blackfin processor in image processing
Images are the main source of human access and exchange of information. Therefore, the application of image processing must involve all aspects of human life and work. As the range of human activities continues to expand, the field of application of image processing will continue to expand. It is spread across aerospace, medical equipment, communications engineering, military security, culture and art. In recent years, with the advent of a series of digital products such as camera phones, digital cameras, digital video cameras, etc., the image signal processing functions contained in these video devices have gradually become recognized and concerned by people. With the rapid development of security products such as digital TVs, IP cameras, web cameras and surveillance cameras in recent years, it shows that the demand for intelligent monitoring is growing, which is the process of continuous enrichment and improvement of image signal processing technology. Among these image signal processing, including auto focus, auto exposure, auto white balance and color conversion, backlight compensation, etc. are some basic image processing algorithms.
Traditional image signal processing is done by a dedicated image processor. With the continuous development of CMOS technology, there has also been a single-chip product that incorporates image signal processing (ISP) in CMOS image sensors. These sensor products can perform some image signal processing algorithms with their integrated functional modules. These basic algorithms include auto exposure, auto white balance, and color conversion. The emergence of such ASIC products was once controversial, because although the ISP built into the sensor has the advantages of reduced power consumption and space saving, but in terms of mobile applications, the ISP is integrated on the sensor in terms of cost. The cost is much higher than the ISP integration on the mobile multimedia processor or baseband. As sensor resolution increases, more and more sensors will only integrate the circuitry necessary to output digital signals, while integrating image processing and compression into a mobile multimedia processor or baseband. From the perspective of image quality, it is better to integrate powerful image processing functions in mobile multimedia processors, and this trend will become more and more obvious with the increase of resolution. So the industry has always had ISP and CMOS sensors that will deviate from the trend of integration prediction. But from the new product line launched by CMOS sensor suppliers in recent years, this prediction has not been fulfilled. Because we have seen, a growing number of CMOS sensor suppliers such as Aptina and OmniVision have introduced the latest CMOS image sensor single-chip products with built-in ISP, and are constantly improving their functions and expanding their application areas. In addition to various image processing functions including exposure control, gamma correction, white balance, color saturation adjustment, and grayscale control, this single-chip product has been introduced to provide excellent low-light performance in night vision. The environment also has excellent near-infrared response performance, which can provide camera manufacturers with more perfect automatic white balance and automatic exposure, and the image performance is not inferior. A series of excellent performances have gradually developed from the original camera phone, digital camera, digital camera application to the field of emerging security products such as web cameras and surveillance cameras. The expansion trend continues, and there is a great momentum.
It should be said that in the pursuit of small and thin consumer products such as digital cameras, the built-in ISP CMOS sensor has certain competitive advantages, low power consumption and small area performance and cost reduction with the scale of production make it a multi-application manufacturer. The darling. At the same time, we have to see that this ISP sensor solution with built-in ISP can only complete some less complicated algorithms, and with the continuous improvement of intelligent requirements, consumer products, especially video surveillance and other security products are The requirement for higher video processing functions requires more complex algorithms, while complex algorithms place higher demands on the processing power of the main chip. For example, many surveillance cameras are based on the standard resolution (CIF) of NTSC (352×240) or PAL (352×288), and the compression format is mainly based on the less complex M-JPEG format. Therefore, the processing performance of the DSP is not high. But with the rapid development of the video surveillance market, people are expecting higher image clarity and higher frame rates. In terms of resolution, the industry expects D1 (704 × 576) or higher resolution; in terms of codec, it is expected to adopt a complex MPEG-4 compression format. At the same time, with the development of network interconnection, when DVR, DVS or IP cameras are used as network terminals, it is also necessary to consider how to occupy less bandwidth during transmission. Therefore, MPEG-4 is moving to H.264 with higher compression ratio. The AVS encoding format evolves. These more advanced video signal resolutions and compression formats will greatly increase the processing power requirements of the core processor. In addition to the powerful processing power, the processor platform is required to provide sufficient flexibility. From the current point of view, although H.264 is the most popular in the compression format, it is expected to become the mainstream in the next few years. However, due to the advantages of codec algorithms such as M-JPEG, H.263, MPEG4 and H.264, coupled with the diversification of market demand, these standards will coexist for some time. To make the program competitive, it is necessary to support as many standards as possible. At the same time, media processing formats and algorithms are constantly being updated. When more advanced algorithms emerge, engineers hope to simply upgrade software-based rather than redesign based on the original solution. At the same time, video surveillance has its special requirements for security technologies such as confidentiality. Whether it is commercial or home applications, the device itself has the threat of being hacked and hacked. To ensure the security of the device, the product design is undoubtedly A big challenge. At this time, the built-in ISP's single-chip CMOS sensor products showed the inherent shortcomings, because they could not complete such complex functions, but a separate image processor was required to meet the requirements. Although ASIC has cost advantages, it has inherent disadvantages in dealing with media formats and algorithms. In the video surveillance market where audio and video standards continue to evolve and algorithms are constantly improving, ASICs are inferior.
Here, traditional processor vendors such as ADI, TI, NXP, etc. have irreplaceable competitiveness. The new embedded processor Blackfin series, designed by Analog Devices to meet the computational requirements and power constraints of today's embedded audio, video and communications applications, is particularly eye-catching for its outstanding performance.
The Blackfin family of processors has unique technical advantages in image processing, including:
Upgrade to the latest video standards to provide users with the ability to handle all existing standards and future standards;
Dynamic power management. All Blackfin processors feature a variety of power saving technologies. The Blackfin processor is based on a gated clock core design that selectively shuts down the power to the functional unit on a code-by-order basis. It also supports a variety of power-down modes during periods when there is little or no CPU action required. The processor also supports a self-contained dynamic power management circuit with which the operating frequency and voltage can be independently controlled to meet the performance requirements of the executing algorithm. These conversions can occur continuously under the control of an RTOS or user firmware. Most Blackfin processors offer an on-chip core regulator circuit that operates from voltages as low as 0.8V, making them ideal for portable applications that require extended battery life.
Has a high performance processor core. The processor architecture is based on a 10-level RISC MCU/DSP pipeline and a hybrid 16/32-bit instruction set architecture designed for optimal code density, which is ideal for high-density, high-volume image processing systems;
High bandwidth DMA. The image processing system needs to process a large amount of image data, which involves memory data access. Blackfin's two-dimensional DMA controller can realize automatic data transmission, greatly reducing the processor's kernel overhead; dedicated video instruction set, image There is a large amount of video pixel processing in the processing. The Blackfin processor has an instruction set specially designed for 8-bit video data and common video pixel processing algorithms, which greatly improves the processing speed;
Graded storage space. Blackfin has two levels of Cache for L1 and L2. For some data that is not suitable for DMA operations, data/code buffering mode can be used to speed up processing.
Rich peripheral functions. Blackfin has a high-speed serial/parallel data interface that seamlessly connects to other high-speed devices;
In terms of security, Blackfin Lockbox Secure Technology uses a combination of hardware and software to implement various security features. The public key that authenticates the system is stored in a user-programmable area in single-programmable (OTP) memory that can be controlled and defined by the OEM. At the same time, because of its programmability, the instruction set can implement rich software encryption algorithms, which means that the same device can support multiple content protection formats. Therefore, the technology can provide a secure environment for the entire video surveillance device from the aspects of source verification, content integrity, confidentiality and updateability;
Low power consumption and low cost are also the keys to Blackfin's outstanding price/performance ratio.
Based on the above technical features, and because of its powerful multimedia processing capabilities, professional algorithm support, and embedded real-time operating system, Blackfin can not only achieve excellent autofocus and other 3A1B algorithms, but also perform complex such as automatic selection of codes. Computationally intensive algorithms, especially in products such as intelligent video surveillance where processor requirements are extremely high.
Taking intelligent video surveillance products as an example, although great progress has been made, there is currently no recognized best method in the field of intelligent video analysis. The complexity of its own research content makes the algorithm complex, the scope of application is limited, the research methods and tools are diverse, and there is no universal method that meets the requirements of robustness, accuracy and speed. At the same time, the requirements of networked and distributed processing of video surveillance systems, as well as the limitations of cost, size and power consumption in large-scale engineering installations, make embedded processors with increasing computing power and bandwidth become the mainstream of video surveillance systems. select. Instead of standardized intelligent video analysis, it is the use of DSP.
Blackfin's dozens of DMA channels and flexible configuration Caches are well suited to the high computational throughput and high data throughput requirements of video surveillance systems. The 10-stage pipeline gives Blackfin a strong command parallel execution capability. The zero overhead loop control instruction causes a large number of loop jumps in the system to no longer consume any processor clock cycles. Taking advantage of these advantages, the real decoder's idct4×4 algorithm is 7 times faster on Blackfin.
In addition, video data has its own characteristics. Within a different color space, each component representing a pixel is typically 8 bits wide. Blackfin's four video arithmetic units and video pixel instruction sets greatly speed up video operations. A video pixel operation instruction can complete 11 kinds of video pixel operations such as addition, subtraction, addition and subtraction mixing, averaging or subtraction and absolute value of 4 pairs of video data components in one cycle. These operations are widely used in motion estimation, loopfilter, and various algorithms for intelligent video analysis in codec algorithms. In some basic operators of intelligent video analysis, such as histogram statistics, median operations, Sobel operations, expansion operations in morphology, etc., Blackfin's MIN, MAX instructions can be used to eliminate conditional jumps and save processor cycles. Not only that, Blackfin also supports vector operations for 13 non-video data. Properly design the data structure, Blackfin's special instructions can be used in the previous background separation, threshold calculation and update to make the intelligent video analysis algorithm faster. Most of these valid instructions can be executed in parallel, doubling Blackfin's processing power.
On the application side, ThinkSmart, a leading intelligent video surveillance device based on the Blackfin series' latest product BF561, developed by Nanjing Xintiantian Technology Co., Ltd. and the University of Hong Kong, has been introduced to the market. The BF561's unique dual-core design enables ThinkSmart to independently run image processing and system control tasks on the processing core to achieve optimal overall performance, enabling this product to easily and efficiently detect threats or random visitors, enabling a range of intelligence Video surveillance.
The Blackfin family of processors is ideal for a wide range of applications in image processing, especially in terms of intelligent technical challenges. ADI is also working to improve the performance of the Blackfin family and increase its price/performance ratio through the latest CMOS deep submicron process scale, deeper pipeline, multi-core, more computing units per core, and full use of signal processing acceleration engines. And the leading position in performance and power consumption ratio.