An image-based document translation system consists of several components, among which OCR (Optical Character Recognition) plays an important role. However, existing OCR software is not robust against environmental variations. Furthermore, OCR errors are often propagated into the translation component and cause, causing poor end-to-end performance. In this paper, we propose an imagebased document translation using an error correction model to correct misrecognized words from OCR output. We train our correction model from synthetic data with different fonts and sizes to simulate real world situations. We further enhance our correction model with bigrams to improve our word segmentation error correction. Experimental results show substantial improvements in both word recognition accuracy and translation quality. For instance, in an experiment using Arabic Transparent Font, the BLEU score increases from 18.70 to 33.47 with the use of our noisy channel model.

The reading of characters by computer is known as Optical Character Recognition (OCR). OCR is one of the popular applications of image processing systems. OCR has many different practical applications. This paper is to develop the handwritten Myanmar alphabet recognition (HMAR) system which is able to evaluate the performance of Handwritten Optical Character Recognition (HW-OCR). This HMAR system is able to recognize handwritten character of several different writing styles. This paper presents the feature extraction method for the handwritten Myanmar alphabet based on the zoning method. The development of rule-based recognition system for Myanmar alphabet is supported as a result to obtain better recognition accuracy rate. 1.

The sketch community has, over the past years, developed a powerful arsenal of recognition capabilities and interaction methods. Unfortunately, many people who could benefit from these systems lack pen capture hardware and are stuck drawing diagrams on traditional surfaces like paper or whiteboards. In this paper we explore bringing the benefits of sketch capture and recognition to traditional surfaces through a common smart-phone with the Sketch Practically Anywhere Recognition Kit (SPARK), a framework for build-ing mobile, image-based sketch recognition applications. Naturally, there are several challenges that come with recognizing hand-drawn diagrams from a single image. Image processing techniques are needed to isolate marks from the background surface due to variations in lighting and surface wear. Further, since static images contain no notion of how the original diagram was drawn, we employ bitmap thinning and stroke tracing to transform the ink into the abstraction of strokes commonly used by modern sketch recognition algorithms. Since the timing data between points in each stroke are not present, recognition must remain robust to variability in both perceived drawing speed and even coarse ordering between points. We have evaluated Rubine’s recognizer in an effort to quantify the impact of timing information on recognition, and our results show that accuracy can remain con-sistent in spite of artificially traced stroke data. As evidence of our techniques, we have implemented a mobile app in SPARK that captures images of Turing machine diagrams drawn on paper, a whiteboard, or even a chalk-board, and through sketch recognition techniques, allows users to simulate the recognized Turing machine on their phones.

Malayalam, the official language of Kerala, a southern state of India has been accorded the honour of language of eminence. Hence the researches in recognition and related works in Malayalam language is gaining more prominence in the current scenario. This paper proposes the use of Curvelet transform and neural network for the recognition of handwritten Malayalam character. Curvelet transform is to be used in the feature extraction stage and neural network for classification. Curvelet transform provides a compact representation for curved singularities and is well suited for malayalam language. Two different back propagation algorithms had been employed and the performance is compared on varying architecture. The promising feature of the work is successful classification of 53 characters which is an improvement over the existing works. Application of character recognition include sorting of bank cheques and postal letters, reading aid for blind, data compression etc. Besides, an automated tool with graphical user interface in MATLAB has been developed for Malayalam character recognition.

Thermal imaging has been used in the past for remote detection of regions of canopy showing symptoms of stress, including water deficit stress. Stress indices derived from thermal images have been used as an indicator of canopy water status, but these depend on the choice of reference surfaces and environmental conditions and can be confounded by variations in complex canopy structure. Therefore, in this work, instead of using stress indices, information from thermal and visible light imagery was combined along with machine learning techniques to identify regions of canopy showing a response to soil water deficit. Thermal and visible light images of a spinach canopy with different levels of soil moisture were captured. Statistical measurements from these images were extracted and used to classify between canopies growing in well-watered soil or under soil moisture deficit using Support Vector Machines (SVM) and Gaussian Processes Classifier (GPC) and a combination of both the classifiers. The classification results show a high correlation with soil moisture. We demonstrate that regions of a spinach crop responding to soil water deficit can be identified by using machine learning