I guess I know what you mean. You want to calculate the temperature of an image based on intensity values. You have started in a right way what you need is to take the measured temperature as the target and the intensity as an input to an ANN model then you will be able to use the created model to obtain temperature every time you provide it with intensity. This concept is applied with remote sensing (thermal bands) some people create a daily sea temperature map from a daily thermal band by creating a regression model using measured temperature value as you did. So, instead of doing measurement every time the temperature map is created at once. 

You need to create an ANN model or linear regression model from the data you collected (intensity and temperature) good luck

If you are working already on a set of images and calculating the temperature from the intensity, then you must need the experimental set up. First of all, you have to control the temperature of a room (may be small chamber), then slowly increase the temperature and take the images at all the temperature ranges. Then calibrate the intensity coming out of all the images. Once calibrated, your algorithm is ready to be implemented on other image sets. 

Hope it may help you.

Dear Mr. Kumar,

The question is how did you convert the thermal band (intensity values) into temperature value? what do you mean test image and reference image how did you get them?

Your question needs more info such as what training and test data you used to create your temperature map from thermal image? 

@Mr.M.Awad,

In my project the images were captured using a thermal camera.i have images of an object with various temperatures.for example, the image of the object is captured at 250c ,after that the image of the same object is captured at 300c .after that,40,50,60 like that.after image acquisition and temperature measurement,the mean intensity of each image is calculated.using this values ,I plot a graph(mean intensity of an image vs temperature).this graph shows fine variations in the intensity of the image according to the temperature.After that,i store this mean intensity and corresponding temperature in the data file.these images are used as reference images.Now i want to develop a automatic program,which compares the mean intensity of the test image(thermal image of an image with unknown temp value) and  mean intensitry in the data file,and display the temperature based on this comparison.

@ Mr.Gusharanjeet Singh Kalra,

I completed the experimental section. i need a matching technique to match the intensity of the test image with  more number of reference intensity in the data set and display the corresponding temperature value..

I guess it's better to do a review of related paper. After reading them, you will find a solution.

Your experiment is fine but it will work only and only if you maintain the light intensity of the room same as when the reference images were taken. If you keep the room intensity conditions same, then all the temperatures calculated in n-images will be same. But if you changed the light intensity then the calculations will be wrong.

In this case you have to add another parameter in your experiment for reference images which will be Lux-meter. It will also measure the light intensity at that time. You have to again repeat your reference images with different light intensity.

Now, when you are going to take the image under testing, also measure the light intensity and then your temperature will come out to be exact with minor error.

I guess I know what you mean. You want to calculate the temperature of an image based on intensity values. You have started in a right way what you need is to take the measured temperature as the target and the intensity as an input to an ANN model then you will be able to use the created model to obtain temperature every time you provide it with intensity. This concept is applied with remote sensing (thermal bands) some people create a daily sea temperature map from a daily thermal band by creating a regression model using measured temperature value as you did. So, instead of doing measurement every time the temperature map is created at once. 

You need to create an ANN model or linear regression model from the data you collected (intensity and temperature) good luck

@Mr.Gusharanjeet Singh Kalra , Mr.M. Awad 

Thank you very much for your valuable suggestions on my question..

Dear Mahesh,

it looks like a calibration problem. As M. Awad suggested, you need to look for a suitable regression model. That may be a linear, polynomial or even a more complex one. I'm attaching a link to a related document, you may find it interesting. Hope it helps.

Best,

f.

http://www.nmschembio.org.uk/dm_documents/LGCVAM2003032_xsJGL.pdf

Yes, Histogram matching is possible.it is not necessary to compare the mean value.Just compare the histogram i.e, the probability of occurrence of definite intensities in the specified range of intensities

@ Mr.Francesco bianconi

thank you for your response

Image enhancement is to improve the interpretability or perception of information in the images to provide better input for other automated image processing steps. The image acquired from natural environment with high dynamic range includes both dark and bright regions. Due to exceed in dynamic range of human eyes sensing, those images are difficult to perceive by human eyes. Image enhancement is a common approach to improve the quality of those images in terms of human visual perception. Enhancement techniques can be divided into two categories namely:

1. Spatial domain methods

2. Transform domain methods.

Histogram Equalization is a technique that is used to enhance the image contrast by basically stretching the dynamic range of the image. Histogram Equalization is a technique that generates a gray map which changes the histogram of an image and redistributing all pixels values to be as close as possible to a user - specified desired histogram. HE allows for areas of lower local contrast to gain a higher contrast. Histogram equalization automatically determines a transformation function seeking to produce an output image with a uniform Histogram