A low-overhead adaptive image compression technique for energy-constrained WMSN

Highlights

• Concept of ‘rough segment’ is introduced based on image segmentation and set theory.

• Multilayered template is generated off-line using the proposed rough segment.

• Based on the above template image segmentation is performed.

• Image is compressed by dropping columns of image pixel array adaptively to save energy.

• Comparative performance of the proposed scheme is evaluated both through qualitative and quantitative analyses.

Abstract

Advancement of low-cost hardware capable of processing multimedia data coupled with efficient short range radio communication has led to the development of Wireless multimedia sensor network (WMSN). We consider post-disaster situation analysis application where the nodes in a WMSN capture disaster images from the affected area. Energy is one of the scarcest resources for such a network especially for disaster application where power supply for charging the battery of a node is nearly impossible. In this paper we exploit such facts and propose an energy saving multimedia (image) data processing scheme for its subsequent transmission over energy-constrained WMSN. Unlike conventional image processing, it saves energy by reducing the data to be transmitted at the cost of segmentation-less adaptive compression. Finally, the comparative performance of the scheme shows its dominance over all the competing schemes. Precisely, our scheme saves on an average 34% energy over all the competing schemes.

Introduction

The Wireless multimedia sensor network [1] are capable of processing multimedia data like image, video and audio streams. A wide range of applications has come up exploiting the multimedia capabilities of the nodes. The applications vary from continuous monitoring application (e.g. post-disaster situation analysis) to event driven application (e.g. forest fire detection). In this work we consider the application post-disaster situation analysis through capturing images of the affected area. In such an application, it is essential to collect visual information of the affected areas through the capturing of images of the same by the nodes of the network and transmitting the data wirelessly through multiple hops towards high end receiving node (e.g. sink) that reconstructs the captured images. With the help of such visual information severity of destruction may be assessed and accordingly rescue and relief operation may be planned.

Energy is one of the scarcest resources in such networks, especially it is scarce in transmitting multimedia data. For a sensor node, energy consumption during communication [2] is much more than that of data processing inside it. Moreover, in a disaster affected area due to unavailability of power supply, battery replacement of nodes is nearly impossible. Therefore, it is essential to save energy during transmission. Reducing the size of the images to be transmitted is one of the solutions for the same. If data can be minimized using a lightweight compression technique before transmission, then overall energy consumption can be reduced. Moreover, from viewers’ perspective entire image may not be equally important. If the degree of compression is applied adaptively on a region of an image based on its importance within the image, then for a less important region higher compression ratio may be applied and that reduces the volume of transmitted data thereby ensures further energy saving. This approach requires dividing the image into several regions by appropriate image segmentation algorithm. However, applying segmentation algorithm directly in tiny mote incurs too much computational overhead that in turn draws energy of the battery of the mote. Therefore, alternative mechanisms for segmentation need to be devised to implement the said adaptive compression technique.

Many works have been done so far in this area. Most of these works require either additional hardware or incur intensive computation or significant amount of memory. Neither of these solutions is affordable for the energy-starved, resource-constrained nodes in WMSN. Further, in some of the WMSN based applications (e.g. post disaster situation analysis) it is rather important to save energy of the nodes as much as possible while reconstruction quality may be compromised to a certain extent. This motivates us to explore a solution towards the development of such low-overhead, energy-saving image processing scheme without requiring additional hardware or memory for the said application while keeping reconstruction quality within an acceptable limit.

The main contributions of this paper are as follows:

• We introduce a new concept of rough segment based on conventional image segmentation and set theory. Based on the proposed rough segment, application oriented template with multilayered segment is generated off-line.

• Based on the above template image segmentation is performed.

• We propose an energy saving image compression approach where columns of the image pixel array are dropped adaptively for different segments of an image.

• The performance of the scheme is evaluated both through qualitative and quantitative analyses. A comprehensive comparative performance in terms of different overheads including energy overhead is also provided.

The paper is organized as follows. Literature survey is provided in Section 2. In Section 3, we introduce a new concept, rough segment along with its features and characteristics. Section 4 presents the rough segment based proposed offline template generation scheme for segmentation. A template based adaptive image compression scheme is presented in Section 5. In Section 6, the performance of the scheme is evaluated based on qualitative analysis and quantitative results. Finally, in Section 7, the paper is concluded with some mention about the future scope of the present work.