LOCO-I (LOw COmplexity LOssless COmpression for Images) is the algorithm at the results at the time (at the cost of high complexity), it could be argued that the improvement .. In the sequel, we assume that this term is tuned to cancel R. LOCO-I (LOw COmplexity LOssless COmpression for Images) is the . Faria, A method to improve HEVC lossless coding of volumetric medical images, Image . A. Lopes, R. d’Amore, A tolerant JPEG-LS image compressor foreseeing COTS. Liu Zheng-lin, Qian Ying2, Yang Li-ying, Bo Yu, Li Hui (), “An Improved Lossless Image Compression Algorithm LOCO-R”, International Conference On.

Author: Kimuro Mooguzilkree
Country: Kazakhstan
Language: English (Spanish)
Genre: Personal Growth
Published (Last): 28 February 2017
Pages: 114
PDF File Size: 16.12 Mb
ePub File Size: 9.12 Mb
ISBN: 915-1-52273-229-5
Downloads: 93198
Price: Free* [*Free Regsitration Required]
Uploader: Dimuro

In this paper, a new low complexity and lossless image compression system for capsule endoscopy CE is presented. The compressor consists of a low-cost YEF color space converter and variable-length predictive with a combination of Golomb-Rice and unary encoding.

There was a problem providing the content you requested

All these components have been heavily optimized imprived low-power and low-cost and lossless in nature. As a result, the entire compression system does not incur any loss of image information. Unlike transform based algorithms, the compressor can be interfaced with commercial image sensors which send pixel data in raster-scan fashion that eliminates the need of having large buffer memory.

Finally, a complete capsule endoscopy system is developed on a single, low-power, nm field programmable gate arrays FPGA chip. The prototype is developed using circular PCBs having a diameter of 16 mm. Several in-vivo and ex-vivo trials using pig’s intestine have been conducted using the prototype to validate the performance of the proposed lossless compression algorithm. The results show that, compared with all other existing works, the proposed algorithm offers a solution to wireless capsule endoscopy with lossless and yet acceptable level of compression.

Capsule endoscopy CE [ 12 ] is a non-invasive technique to receive images of the intestine for medical diagnostics. The main design challenges of endoscopy capsule are acquiring and transmitting acceptable quality images wlgorithm utilizing as little hardware and battery power as possible. In order to save wireless transmission power and bandwidth, an image compressor needs to be implemented inside an endoscopy capsule.

Lossy image compressors produce some difference between the original and reconstructed images. For medical diagnostics, the distortion of the reconstructed image can lead to inaccurate diagnostics decisions, though in medical and endoscopic imaging, lossy compression is acceptable up to a certain point for example, a compression ratio of 15 was found as the visually lossless threshold for the JPEG lossy algorithm [ 3 ].

However, in hospitals in improced days where Picture Archiving and Communication Systems PACS are used to store medical and algoritgm data in digital form, lossless compression is a requirement [ 4 ]. In addition, lossless compressors produce identical reconstructed images compared with the original images without any distortion.

Therefore, the prime objective has always been to find efficient and lossless compression methods. In this paper, a low complexity yet lossless image compression algorithm is presented that is purposely designed for capsule endoscopy.

The compressor works on a highly efficient YEF color space [ 5 ], which is specially designed to compress endoscopic images by analyzing the aan image properties. After color space conversion, the compressor takes the difference of a pixels using differential pulse coded modulation DPCM and then encodes the differences in variable length Golomb-Rice [ 6 ] and unary coding. A customized corner clipping scheme is also implemented to remove uninteresting corner area of the image to increase compression ratio.

All these components are fully optimized for low-cost operation and lossless in nature; as a result, no loss of any diagnostic information takes place inside the endoscopic system. In order to validate the performance of the compression algorithm, it is deployed inside an endoscopic capsule prototype developed in our lab. At first, a modular and programmable CE development system platform consisting of a miniature field programmable gate array FPGA based electronic capsule is developed.

The prototype supports various imaging modes including the commonly used white light imaging WLI and narrow band imaging NBI [ 7 ], and communicates with the data logger in full duplex fashion, which enables configuring the image size and imaging mode in real time during the examination.

The CE prototype is then tested to assess the performance using live pig in both ex-vivo and in-vivo trials at the animal facility. Both lossy and lossless image compression algorithms are found in the literature targeting capsule endoscopy application. A brief discussion on both types of compression is given below. The lossy algorithms found in literature are mainly based on transform coding where the Discrete Cosine Transform DCT is used [ 8 — 15 ].


However, commercially available complementary metal-oxide-semiconductor CMOS image sensors [ 1617 ] send pixels in raster scan fashion i. These image sensors also do not have internal buffer memory for image storage and random access of pixels. Due to the mismatch of pixel steaming sequence of commercial image sensors and pixel access sequence required by transform based compression algorithms, buffer memory needs to be implemented inside the capsule to store a complete or blocks of an image frame, so that the image pixels can be accessed by the compressor in block wise fashion from the buffer memory.

Pixel access sequence in image sensor using two topologies: The memory requirement in bits, Scan be expressed as Equation One solution of this problem is to pause the image sensor after it sends all the pixels for accessing a block; however the feature of pausing an image sensor in the middle of transmitting pixels of a frame is not found due to the read out timing requirements of the pixel array in the image sensors.

Another possible solution to the problem is to use two buffer memory of total size 2 Sso that while the compressor works with pixels of one buffer, the new pixels continuously coming from the image sensor are stored in the other buffer.

However, this solution can create timing error if compression and transmission time exceeds the input data rate of the image sensor. A safer solution of this problem could be to use sufficient memory to store a complete image frame. Buffer memory takes significantly large silicon area and consumes sufficient amount of power which can be a noticeable overhead in capsule endoscopy application. Moreover, complex calculations are associated with transform coding based compressors i.

The works in [ 518 ] by our group present lossy compression algorithms which do not require block based access of image pixels; rather they can work with pixels coming in raster scan fashion. However, these compressors both raster scan based and block based presented in the literature are lossy compressors which incur various levels of distortions in the reconstructed images.

For medical diagnostics, these distortions can lead to inaccurate diagnostics decisions. Moreover, these works present only computer based simulation results using endoscopic images taken from on-line databases such as [ 19 ]. The performance of their algorithms was not validated using hardware level simulation or ex-vivo trials.

While lossy compression algorithms for capsule endoscopy are in abundance, their lossless counterpart is only a few. In [ 20 ], our group proposed a lossless image compressor based on YUV color space.

However, YUV color space is computationally expensive due to the presence of floating point numbers during conversion from RGB and imagee consumes significant area imaeg power when implemented in hardware. Besides, the practical limitation of the work is that the architecture assumed that the image sensor must have built-in RGB to YUV color space converter which limits the robustness. Moreover, the performance was validated losslfss only image simulation without any deployment into hardware platform or any capsule endoscopic system.

No work was found that uses JPEG-LS for capsule endoscopy application; as a result, we have implemented it and applied to our dataset. The results are added later in this paper. Lastly, the work in [ 24 ] proposed a modified JPEG-LS algorithm for endoscopic image compression which is not entirely lossless. The work provides simulation results only without any in-vivo trial for performance validation. In contrast, here the proposed lossless compressor can work with any elementary RGB based image sensor that outputs image pixels in raster scan fashion which eliminates the need of a memory buffer.

The compressor consists of a low complexity color space, known as YEF, designed by analyzing the unique properties of endoscopic images [ 5 ]. As the compressor losslees lossless, it produces reconstructed images without any distortion and thereby reduces the possibility for inaccurate diagnostics.

The performance is validated using a complete CE system developed in our lab with ex-vivo trials with live pig. While designing the lossless compression algorithm, we have set the following design objectives:.

The block diagram kmproved the proposed lossless compression algorithm is shown in Figure 2.

Then an optional clipping module is added that removes uninteresting corner area of the image. A lossless predictive encoder, known as differential pulse coded modulation DPCM is used. The differential values of luminance component are encoded in Golomb-Rice code [ 6 ] where the differential values of chrominance components are encoded in unary code. The different stages of the proposed algorithm as placed in the processing pipeline are briefly discussed below:. At this first stage of the algorithm, RGB pixels are converted to YEF color space [ 5 ], which is suitable for CE image compression and efficient for hardware implementation.


The color space is designed by analyzing the unique properties of endoscopic images for better compression. The motivation for the YEF color space comes from the fact that, endoscopic images generally exhibit dominance in red color with the absence of significant green and blue components.

Our experiments show that, in most cases, the intensity distribution of green in endoscopic images is very similar to that of blue component. Experiments have also shown that the intensity distribution of luminance Y has similar pattern of green and blue components — thus, subtracting green and blue components form the luminance will produce differential pixel values of almost equal numbers and will reduce the entropy of the chrominance planes. Reduced entropy will cause higher compression ratio in the chrominance planes.

In YEF, the luminance is stored in Y component, E stores the difference between luminance and green component, and F stores the difference between luminance and blue component. The relationships are shown in Equations 2 — 4.

Marcelo Weinberger – Google Scholar Citations

It can be seen from Figure 4 that, after converting to YEF, there is less change in pixel values in chrominance E and F components of YEF color space than RGB components, which indicates that less information or entropy is contained lodsless and these two components can be compressed heavily.

In Figure 5the intensity distributions for the NBI image Figure 3b are shown and they also reveal that the chrominance components of YEF color space contain low information content or entropy. Intensity distribution of color components of a WLI image: Intensity distribution of color components of an NBI image: Note that, the YEF color space does not discard the chrominance information; in fact, it is another representation of the RGB color space cimpression is more suitable for compression and theoretically lossless.

From Equations 2 — 4it is observed that the conversion between color spaces involves only a few additions and divisions by numbers that are powers of 2, which can be implemented by shift operations in digital hardware. When Equations 2 — 4 are implemented in digital hardware as integers, minor variations in the pixel values may occur due to the rounding of fractions to integers. The YEF color space can be made fully reversible i. In capsule endoscopy, the corner areas in a captured image are often blacked out.

To achieve it, an optional corner clipping algorithm can be employed as described in [ 20 losxless during the image acquisition stage. In this stage, a lossless predictive coder imabe used. Due to clmpression rare occurrence of sharp edges in endoscopic imae, the difference between the component values of two consecutive pixels is generally small.

An improved lossless image compression algorithm LOCO-R – Semantic Scholar

The change in component values dX with respect to its adjacent left pixel in any row is given by Equation algorifhm X imoroved represent Y, E, or F component values. From Figure 6it is seen that smaller changes in pixels values occur in endoscopic images. Similar phenomenon is observed for other endoscopic images used in this work. More simulations have been conducted with WLI and 15 NBI test endoscopy images and with standard images; average absolute difference AAD is used as given in Equation 9 as the statistical measure of dX:.

The results are summarized in Table 2where it is seen that, in general, the difference in pixel dX with respect to the adjacent left pixel is very small in endoscopic images compared to that of standard images. As a result, the DPCM is a good choice. It should be noted that, the DPCM used here does not use any quantization and thus is lossless as well.

Besides, it has very zn computational complexity which will help reduce power and area consumption of the compressor. The difference of the consecutive imahe dX is then mapped to a non-negative integer and then they are encoded in variable length coding.

To get the best compression ratio, the difference of luminance dY is fompression in Golomb-Rice code and the difference of alogrithm dE and dF are encoded in unary for WLI images. In Golomb-Rice coding, the choice of k parameter is important since it dictates the code length. A detailed discussion on choosing k parameter can be found in Section 3. This phenomenon can be noticed from Table 2. The k parameter values are summarized in Table 3.