iOS Black and White Image and C API

[tomcontr]

Hi guys,

I'm trying to create a code that allows me to clear images to be read by and OCR engine. So far I haven't been able to accomplish much. Only install Imagemagink on my iOS project and use some of the API functions to create a Gray Scale image.

What I'm trying to achieve is something like this: http://www.fmwconcepts.com/imagemagick/textcleaner/

But I'm not too familiarized with the API. Does anyone knows how to do something like that using the API?

This is my code by the way.

regards,

Code: Select all

-(UIImage *)drawMonochromeImage:(UIImage *)image
{
    // Create temporary file
    NSString *tempFilePath = [NSTemporaryDirectory()
                              stringByAppendingPathComponent:@"temp.jpg"];
    
    MagickWandGenesis();
    MagickWand *wand = NewMagickWand();
    NSData *data = UIImagePNGRepresentation(image);
    MagickReadImageBlob(wand, [data bytes], [data length]);
    
    // Monochrome image
    //MagickQuantizeImage(wand,2,GRAYColorspace,1,MagickFalse,MagickFalse);
    
    
    MagickDespeckleImage(wand);
    MagickEnhanceImage(wand);
   
    MagickQuantizeImage(wand,256,GRAYColorspace,0,MagickFalse,MagickFalse);
    MagickBrightnessContrastImage(wand,-40,30);

    // Write to temporary file
    MagickWriteImage(wand,
                     [tempFilePath cStringUsingEncoding:NSASCIIStringEncoding]
                     );
    
    // Load UIImage from temporary file
    UIImage *imgObj = [UIImage imageWithContentsOfFile:tempFilePath];
    
    // Display on device
    return imgObj;
   // [self.imageView setContentMode:UIViewContentModeScaleAspectFit];
}

[fmw42]

Use the API equivalent of the command -lat. That is what I used in my script, textcleaner.

[tomcontr]

Thanks!

This is what I´ve found.

Code: Select all

 MagickBooleanType MagickAdaptiveThresholdImage(MagickWand *wand,
    const size_t width,const size_t height,const ssize_t offset)

I will try it and see what comes out.

Regards!

[tomcontr]

Ok, so after diggin around I came up with the code below. The input image and output image are at the bottom:

My goal is to clean the image and have a clean monochrome image which black font and white background.

Hope thant you could give me some pointers,
regards!

Code: Select all

-(UIImage *)drawMonochromeImage:(UIImage *)image
{
    // Create temporary file
    NSString *tempFilePath = [NSTemporaryDirectory()
                              stringByAppendingPathComponent:@"temp.jpg"];
    
    UIImage *image2 = [UIImage imageNamed:@"IMG_2517.JPG"];
    CGImageRef imageRef = [image2 CGImage];
    
    MagickWandGenesis();
    MagickWand *wand = NewMagickWand();
    MagickWand *wand2 = NewMagickWand();
    NSData *data = UIImagePNGRepresentation(image2);
    MagickReadImageBlob(wand, [data bytes], [data length]);
    
    // Monochrome image
    //MagickQuantizeImage(wand,2,GRAYColorspace,1,MagickFalse,MagickFalse);
    
    MagickQuantizeImage(wand,256,GRAYColorspace,0,MagickFalse,MagickFalse);
    
    //MagickDespeckleImage(wand);
    //MagickEnhanceImage(wand);
    
    wand2 = CloneMagickWand(wand);
    MagickNegateImage(wand2,TRUE);
    MagickAdaptiveThresholdImage(wand2,320,103,0); //self.aa);
    MagickContrastImage(wand2,0);
    
    MagickContrastImage(wand,0);
    
    
    MagickSetImageCompose(wand2,CopyOpacityCompositeOp);
    
    
    PixelWand *color = NewPixelWand();
    PixelSetColor(color, "white");
    
    PixelWand *color2 = NewPixelWand();
    PixelSetColor(color2, "black");
    
    
    MagickSetImageBackgroundColor(wand2,color);
    
    MagickCompositeImage(wand,wand2,OverCompositeOp,0,0);
    
    MagickOpaquePaintImage(wand,color2,color,40,true); // this doesnt do anything.
    
    MagickSharpenImage(wand,0.0,1.0);

    
    //MagickBrightnessContrastImage(wand,-40,30);
    
    
    
    // Write to temporary file
    MagickWriteImage(wand,
                     [tempFilePath cStringUsingEncoding:NSASCIIStringEncoding]
                     );
    
    // Load UIImage from temporary file
    UIImage *imgObj = [UIImage imageWithContentsOfFile:tempFilePath];
    
    // Display on device
    return imgObj;
   // [self.imageView setContentMode:UIViewContentModeScaleAspectFit];
}

[fmw42]

try the equivalent of this:

convert test.png -gamma 0.25 -auto-level -negate -lat 30x30+10% -negate result.png

-lat works best if the "object" is white and the background is black when you use the bias term

[tomcontr]

It worked!!

this is the code:


MagickLevelImage(wand,0.0,0.25,MaxRGB);
MagickNegateImage(wand,false);
MagickAdaptiveThresholdImage(wand,30,30,10);
MagickNegateImage(wand,false);

Im almos there. Now I need to do two more things.

1. clear the image a little bit more. For example remove black edges, and sometimes there are small black dots. Is there anything to clean that as much as it can?

2. Is there anything I can do so the numbers stay as much clear as possible. Some times they are a little bit distorted.

Regards!

[fmw42]

1) If the black dots are small, then remove them using -morphology open or close depending upon polarity of the dots. If they are black on white background then use close.

2) Not much that I can think of besides adjusting the contrast, which is what the -gamma did. If they have small holes in them then you can use -morphology open. Often I just do both in series (or use -morphology smooth)

Better resolution images and not using JPG could help further.


see
http://www.imagemagick.org/Usage/morphology/#open
http://www.imagemagick.org/Usage/morphology/#close
http://www.imagemagick.org/Usage/morphology/#smooth

[tomcontr]

sorry for my ignorance, but same thing for the edges?

Regards!

[fmw42]

Sorry, what is the issue with the edges?

[tomcontr]

Some times the edges are black (or shadows) but they are not even on each one of them and that causes the OCR engine to read them as something. What I would like to do is to gently remove them trying to keep what it is in the middle.

For example in the image that I uploaded the white borders would be the black bordes that Im talking about.

[fmw42]

Only way I know would be to shave or chop off those edges. See
http://www.imagemagick.org/Usage/crop/#shave
http://www.imagemagick.org/Usage/crop/#chop

or just crop the part you want
http://www.imagemagick.org/Usage/crop/#crop

[tomcontr]

Ok, I`ve made some progress. I was able to use for the first time the "ConvertImageCommand()" with the API,which I thnk would make things easier.

This code works fine:

Code: Select all

char *argv[] = {"convert", input, "-colorspace", "gray", "-type", "grayscale","-gamma", "0.25", "-auto-level", "-negate", "-lat", "30x30+10%", "-negate", output, NULL};

But This one gives me the erros below:

Code: Select all

char *argv[] = {"convert","-respect-parenthesis","\( ", input, "-set", "-colorspace", "RGB", "-colorspace", "gray", "-type", "-set","colorspace","RGB" "-normalize","\)"," \(","-clone 0","-set","-colorspace", "RGB","-colorspace","gray","-negate", "-lat", "15x15+5%", "-contrast-stretch 0","\)","-compose","copy_opacity","-composite","-fill","white","-opaque","none","+matte", output, NULL};

MyStore: UnableToOpenBlob `( ': No such file or directory @ error/blob.c/OpenBlob/2645.
MyStore: UnableToOpenConfigureFile `delegates.xml' @ warning/configure.c/GetConfigureOptions/705.
MyStore: UnableToOpenBlob `( ': No such file or directory @ error/blob.c/OpenBlob/2645.
MyStore: NoDecodeDelegateForThisImageFormat `( ' @ error/constitute.c/ReadImage/501.
MyStore: UnrecognizedImageType `-set' @ error/convert.c/ConvertImageCommand/2986.

[tomcontr]

Never mind, found the error.
Here is the final code:

Code: Select all

char *argv[] = {"convert","-respect-parenthesis","\(", input, "-set", "-colorspace", "RGB", "-colorspace", "gray", "-type", "grayscale","-colorspace","RGB", "-normalize","\)","\(","-clone", "0","-set","-colorspace", "RGB","-colorspace","gray","-negate", "-lat", "15x15+5%", "-contrast-stretch", "0","\)","-compose","copy_opacity","-composite","-fill","white","-opaque","none","+matte", output, NULL};

[fmw42]

You are using -colorspace RGB, which is linear and not the norm. It makes the image darker. Is that what you really want. Perhaps you should try -colorspace sRGB and see if that is better or worse. Sometimes one or the other works better for what you want to do.

[tomcontr]

There is a tiny difference, but I think RGB serves more the purpose because it gives me a more define image.

Here is the comparison:



Thanks a lot for your help!

regards and merry Christmas!