Mlx90640 infrared thermal imager temperature sensor module development notes (V)

MLX90640 Infrared thermal imager temperature measurement module development notes （ 5、 ... and ） Array interpolation - from 32*24 Pixels to 512*384 Pixels  

MLX90640 Of 32*24=768 Although the pixels are bigger than before 8*8 perhaps 16*8 The pixels have been improved a lot , But if you use these pixels directly, you still can't form a thermal image very well , In order to use these pixels to smooth the image, we need to interpolate them , Use more pixels to draw the image .

Read some other people's algorithms , It feels like polynomial interpolation , Only the combination of interpolation methods is different .  

Algorithm basis  

The representative ones are yangfengjian from Hangzhou University of Electronic Science and Technology 《 be based on MLX90620 Low cost infrared thermal imaging system design 》, Use cubic polynomials + Bilinear interpolation , The original 16*4 Pixels are expanded to 256*64 Pixels . The essence of bilinear interpolation is primary function （ First order polynomial ）. The conclusion of this article is ：  

（1） Bilinear interpolation method requires less computation 、 Fast , But the contrast is low 、 The details are vague .  

（2） Cubic polynomial interpolation , The image effect is clear , High contrast , But it's a lot of calculation .  

（3） First bilinear interpolation, then cubic polynomial interpolation , The effect is better than the above two single interpolation methods .  

（4） First cubic polynomial interpolation, then bilinear interpolation , High and low temperature distribution is more obvious , The image effect tends to be more realistic . meanwhile , This article also uses an evaluation method of image quality --- entropy & Average gradient  

entropy , One of the parameters in thermodynamics to represent the state of matter , Use symbols S Express , Its physical meaning is a measure of the degree of chaos in the system . It is used for image evaluation to indicate the amount of information expressed by the image . The higher the image entropy, the greater the amount of information .  

Average gradient , It refers to the obvious difference in gray level near the boundary or shadow line of the image , That is, the gray change rate is large , The magnitude of this rate of change can be used to express the image sharpness . It reflects the rate of contrast change of small details in the image , That is, the rate of density change in the multidimensional direction of the image , Represent the relative clarity of the image . The higher the value, the clearer the image .

Interpolation implementation  

The first pixel of each row or column is interpolated in front 2 A little bit  

1~n-1 Pixels , Interpolation after each pixel 3 Point last pixel , Interpolate later 1 A little bit  

n+2+(n-1)*3+1=n+2+n*3-1*3+1=4n+2-3+1=4n, namely ： The pixel becomes the original 4 times  

The above processing method , Insert before the first pixel 2 A little bit , Insert after the last pixel 1 A little bit , Next interpolation

when , Should be interpolated before the first 1 A little bit , Interpolation after the last pixel 2 A little bit , To achieve image balance .  

The pixels after each interpolation are the pixels before interpolation 4 times , After twice interpolation , Can be 32*24 Change for 512*384 Pixels . The following is the interpolation algorithm that has been actually used , It's just using Pascal（Delphi） Written , Those who are interested can be changed to

C Linguistic , Just change the corresponding statement , Language is interlinked .  

 

// This is a one-dimensional array interpolation algorithm  

//SourceDatas:TDoubles; One dimensional array before interpolation  

//Dir:Integer; In which direction and at the end 2 It's worth （0： front ;1： At the end of ）  

//times:Integer The number of terms of a polynomial , The first-order polynomial is 2 term , The quadratic polynomial is 3 term  

// Return value ： One dimensional array after interpolation （ The quantity is before interpolation *4） function PolynomialInterpolationArr(  

SourceDatas:TDoubles; Dir:Integer;  

times:Integer):TDoubles;// One dimensional array interpolation  

var  

   i,j,k:Integer;  

   arrCount:Integer;