Low complexity image compression algorithm based on hybrid DPCM for wireless capsule endoscopy
Introduction
Wireless Capsule Endoscopy (WCE) is a non-invasive procedure that uses a capsule size electronic device to extract the pictorial view of Gastrointestinal (GI) tract. Once this device is swallowed by patient, it captures images of GI tract with inbuilt CMOS camera during its journey from larynx to anus. Captured images are transmitted wirelessly to data logger wrapped around the waist of the patient. Received images in data logger are examined by the doctor with the help of Personal Computer [1]. WCE made complete examination of small bowel possible, which is main challenge with conventional endoscopy. The easy and comfortable procedure of WCE as compared to conventional endoscopy has attracted the patients to use it. Accurate diagnosing of disease requires the transmission of best quality (original) images [2]. Wireless transmission of such images are not possible due to limited bandwidth of radio frequency antenna and limited power supply (8–10 hrs) in WCE [3]. In order to transmit a good quality image under these constrains, image compression is an efficient technique. Image compression algorithms are classified in two categories i.e., lossy and lossless. The lossy compression algorithms provide better compression efficiency, but it reduce the image quality and increase hardware cost. However, the lossless compression algorithms provide original image quality at low hardware cost, but they have less compression efficiency. Low computational complexity algorithm will be preferable for WCE due to its limitations. Since lossless image compression algorithm provide original image quality and has low complexity, it is the best choice for WCE.
Many lossless algorithms have been proposed in literature for WCE. CMOS camera of the capsule send the data in Colour Filter Array (CFA) pattern, therefore in [4] lossless standard JPEG-LS has been applied on CFA data. But compression algorithm on CFA data provide less compression efficiency due to non-homogeneity between neighboring CFA pixels. Therefore, to achieve more compression efficiency, transformation and filtration of CFA pixels has been done prior to the compression process in [5]. Implementation of transformation and filtration in real time consumes more power and requires larger area, which are the overheads of WCE. Improvement in technology leads to development of CMOS camera, which provide RGB output format data [[6], [7], [8]]. In [9] JPEG-LS compression algorithm has been applied on YUV colour space instead of CFA data to achieve more compression efficiency. However, prediction modes of Differential Pulse Code Modulation (DPCM) in JPEG-LS are dependent on upper row pixels values to predict the current pixel value. Hence, one row of image need to be stored, which requires extra buffer memory, but it requires extra area and consumes extra power. Moreover, JPEG-LS also required 1.9 kb of additional register array to store contexts and parameters. The high computational complexity of prediction modes in JPEG-LS increases the hardware cost, which is limited in WCE Furthermore, lossless image compression algorithm based on 1D DPCM has been applied on YEF colour transform [10]. In 1D DPCM, extra buffer memory is not required as previous pixel value is considered as prediction value, which leads to low computational complexity. However, it has been observed that DPCM provide less compression efficiency for sharp images where sudden change in colour and structure is observed. To achieve high compression efficiency for sharp image, methods like hierarchical prediction [11], template matching [12] and context matching [13] has been proposed. The high computational complexity of these methods require high memory and consume more power, which is limited in WCE. The endoscopic images generally contains soft images. But sharp images are often found in WCE images due to inadequate lighting system, abnormal moment of capsule and reflection of light through liquids on skin. In the stomach and small bowel regions, images with acids, bubble, disease, worms and undigested material also contains sharp edges. For transmission, difference error value obtained from DPCM is encoded into binary form using entropy coding. An entropy coder with single context adaptive Golomb-Rice (G–R) encoder has been proposed [14]. However, the performance of G–R code degrade for large difference error values, which is often found in sharp images. Moreover, extra memory is needed to regularly update K parameter, which consume extra power and increases hardware cost.
In this paper, a low computational complexity lossless image compression algorithm is proposed for WCE. The algorithm is based on YEN colour space, Hybrid DPCM and Golomb family encoders. In the proposed scheme, early RGB colour image is transformed into new YEN colour space using RGB-YEN colour transform, which is derived based on special properties of endoscopic images. After that, YEN components are compressed by using hybrid DPCM. Hybrid DPCM is a combination of DPCM and adaptive threshold DPCM. Finally, Y component is encoded using signed G–R code, whereas other components such as E and N are encoded using G–R code.
Rest of the paper is organized as follows: Section 2 described the basic concepts of predictive codes scheme. In Section 3, building blocks of proposed algorithm are described. Proposed algorithm is demonstrated and analyzed in Section 4 and 5. Section 6 concludes the paper.
Section snippets
Differential pulse code modulation
Traditional lossless compression algorithm JPEG-LS is based on predictive coding i.e. DPCM [9]. DPCM in JPEG-LS has 7 prediction modes, which are used to predict the accurate current pixel value, which is predicted from the neighboring informative pixels. The predicted value is then subtracted with the original pixel value to produce a difference error value. Difference error values obtained from seven different prediction modes are compared and lowest among them is chosen for further process.
Analysis of endoscopic image
For analysis purpose first of all 200 endoscopic image samples, taken at 20 different positions of GI tract from larynx to anus has been considered. By analyzing these samples, it is observed that
- 1
Red colour dominate over green and blue colour
- 2
Green and blue colour shows almost similar pattern.
For example an endoscopic image of Ileum (small bowel) is shown in Fig. 3a. Fig. 3b represents its colour intensity distribution and Fig. 3c represents the pixels intensity value of different colour
Proposed algorithm
Based on the analysis of endoscopic images, an efficient image compression algorithm is constructed for WCE images. The proposed algorithm is divided into three steps i.e. RGB-YEN colour transform, hybrid DPCM and Golomb family encoders as shown in Fig. 8. YEN colour space is designed based on special properties of endoscopic images, where RGB colour image is transform into YEN colour space. Y, E and N components leads to intra spectrum redundancy, which help in reducing information. Hybrid
Performance analysis
The proposed algorithm is applied on 200 test images of whole GI tract and its performance is measured. The test images are chosen based on amount of time spent by capsule in GI tract and some disease oriented images are also included. The performance of compression algorithm is calculated using two parameters, i.e. the compression ratio (CR) and peak signal noise ratio (PSNR) numerically. The CR and PSNR is given asand
Conclusion
A new hybrid DPCM based image compression algorithm for WCE application has been proposed. First of all, new YEN colour space is created for WCE, which is based on the special properties of endoscopic images. RGB colour image is transformed into YEN colour space using RGB-YEN colour transform, which results in better compression efficiency. After conversion hybrid DPCM compression technique is applied, which is a combination of DPCM and AT-DPCM predictive codes. It is simple predictive coding,
References (19)
- et al.
Lossless compression of HDR color filter array image for the digital camera pipeline
Signal Process. Image Commun.
(2012) - A. Moglia, A. Menciassi, P. Dario, Recent Patents on Wireless Capsule Endoscopy, 2008....
- et al.
Capsule endoscopy: from current achievments to open challenges
IEEE Trans. Biomed. Eng.
(2011) - et al.
Hardware-efficient low-power image processing system for wireless capsule endoscopy
IEEE J. Biomed. Health Inform.
(2013) - et al.
A low power digital IC design inside the wireless endoscopy capsule
IEEE J. SolidState Circuits
(2006) - et al.
A wireless capsule endoscope system with a low-power controlling and processing ASIC
IEEE Trans. Biomed. Circuits Syst.
(2009) - Aptina, MT9V011 Image Sensor, (n.d.). http://www.aptina.com (Accessed 5 May...
- Oni Vision, OVM7690 Camera Cube, (n.d.). http://www.ovt.com (Accessed 5 May...
- Toshiba, TCM8230MD Image Sensor, (n.d.). http://www.sparkfun.com (Accessed 5 May...
Cited by (20)
Recent developments in wireless capsule endoscopy imaging: Compression and summarization techniques
2022, Computers in Biology and MedicineDeep chroma prediction of Wyner–Ziv frames in distributed video coding of wireless capsule endoscopy video
2022, Journal of Visual Communication and Image RepresentationCitation Excerpt :Consecutive images have high similarity in content, colour and texture. Many image compression methods are proposed to reduce the pixel redundancy using prediction-based methods [8–12]. These methods are low complexity algorithms that result in low compression performance.
Entropy encoder for low-power low-resources high-quality CFA image compression
2022, Signal Processing: Image CommunicationCitation Excerpt :Therefore, raw CFA image compression is a very active research topic [5,6,9,10,12,19]. Among the recently proposed methods the high-performance, high-quality ones are based on predictive coding [8,17,20], JPEG-LS [6,9,19], the discrete wavelet transform (DWT) [5,21–24], or the discrete cosine transform (DCT) with predictive coding [10]. However, the existing methods are quite memory-intensive.
Distributed video coding based on classification of frequency bands with block texture conditioned key frame encoder for wireless capsule endoscopy
2020, Biomedical Signal Processing and ControlCitation Excerpt :Moreover, the capsule moves due to peristaltic action at a slow speed which results in more duplicate frames and high temporal redundancy in WCE video which can be removed to achieve better compression. However, the existing prediction based compression methods [17–20] uses DPCM, that works in lossless and lossy modes. These methods can only provide a compression ratio (CR) of 1:2 to 1:3.
Low cost image compression algorithm with colour reproduction algorithm for transmitting video frame of capsule endoscopy using low power
2020, Biomedical Signal Processing and ControlCitation Excerpt :The transmission of these form data take more processing time and power. The consumption of extra processing time and power can be reduced by using image compression algorithm, but it requires some hardware and power for compressing each component data [8]. In the article, a new approach is proposed, where both encoder and decoder play vital role.
Low-Power Low-Area Near-Lossless Image Compressor for Wireless Capsule Endoscopy
2023, Circuits, Systems, and Signal Processing