为 OpenCV fisherface 识别器调整图像大小

Question

我正在使用 OpenCV Contrib 框架来执行面部识别。我的问题是，当隔离面孔（使用 for 循环）时，OpenCV 会裁剪测试图像以仅显示面孔（如面孔周围的 40x40 框）。我需要将此图像调整为 3000x4000（我的训练图像就是这个尺寸）。 我的问题是如何将图像（大部分时间为 40x40）调整为 3000x4000。

以下是我尝试过的几个调整大小的函数及其相应的错误信息。

1.

(UIImage *)imageWithImage:(UIImage *)image scaledToSize:(CGSize)newSize{
    UIGraphicsBeginImageContext(newSize);
    [image drawInRect:CGRectMake(0, 0, newSize.width, newSize.height)];
    UIImage *newImage = UIGraphicsGetImageFromCurrentImageContext();
    UIGraphicsEndImageContext();
    return newImage;
}

错误信息...

OpenCV(3.4.0-dev) Error: Bad argument (Wrong shapes for given matrices. Was size(src) = (1,48000000), size(W) = (12000000,1).) in subspaceProject, file /Users/mustafa/Desktop/OpenCVBuild/opencv/modules/core/src/lda.cpp, line 183
libc++abi.dylib: terminating with uncaught exception of type cv::Exception: OpenCV(3.4.0-dev) /Users/mustafa/Desktop/OpenCVBuild/opencv/modules/core/src/lda.cpp:183: error: (-5) Wrong shapes for given matrices. Was size(src) = (1,48000000), size(W) = (12000000,1). in function subspaceProject

(lldb) 

我在 predict 函数（发生崩溃的地方）之前设置了一个断点，我传递给函数的图像似乎是 3000x4000。

screenshot

2.

 (UIImage *)squareImageWithImage:(UIImage *)image scaledToSize:(CGSize)newSize {
double ratio;
double delta;
CGPoint offset;

//make a new square size, that is the resized imaged width
CGSize sz = CGSizeMake(newSize.width, newSize.width);

//figure out if the picture is landscape or portrait, then
//calculate scale factor and offset
if (image.size.width > image.size.height) {
    ratio = newSize.width / image.size.width;
    delta = (ratio*image.size.width - ratio*image.size.height);
    offset = CGPointMake(delta/2, 0);
} else {
    ratio = newSize.width / image.size.height;
    delta = (ratio*image.size.height - ratio*image.size.width);
    offset = CGPointMake(0, delta/2);
}

//make the final clipping rect based on the calculated values
CGRect clipRect = CGRectMake(-offset.x, -offset.y,
                             (ratio * image.size.width) + delta,
                             (ratio * image.size.height) + delta);


//start a new context, with scale factor 0.0 so retina displays get
//high quality image
if ([[UIScreen mainScreen] respondsToSelector:@selector(scale)]) {
    //0.0
    UIGraphicsBeginImageContextWithOptions(sz, YES, 1.0);

} else {
    UIGraphicsBeginImageContext(sz);
}
UIRectClip(clipRect);
[image drawInRect:clipRect];
UIImage *newImage = UIGraphicsGetImageFromCurrentImageContext();
UIGraphicsEndImageContext();

return newImage;
}

错误信息...

OpenCV(3.4.0-dev) Error: Bad argument (Wrong input image size. Reason: Training and Test images must be of equal size! Expected an image with 12000000 elements, but got 9000000.) in predict, file /Users/mustafa/Desktop/OpenCVBuild/opencv_contrib/modules/face/src/fisher_faces.cpp, line 140
libc++abi.dylib: terminating with uncaught exception of type cv::Exception: OpenCV(3.4.0-dev) /Users/mustafa/Desktop/OpenCVBuild/opencv_contrib/modules/face/src/fisher_faces.cpp:140: error: (-5) Wrong input image size. Reason: Training and Test images must be of equal size! Expected an image with 12000000 elements, but got 9000000. in function predict

(lldb) 

这给了我一张 3000x3000 的图片。

3.

- (UIImage *)scaleImageToSize:(CGSize)newSize image:(UIImage *)image {

CGRect scaledImageRect = CGRectZero;

CGFloat aspectWidth = newSize.width / image.size.width;
CGFloat aspectHeight = newSize.height / image.size.height;
CGFloat aspectRatio = MIN ( aspectWidth, aspectHeight );

scaledImageRect.size.width = image.size.width * aspectRatio;
scaledImageRect.size.height = image.size.height * aspectRatio;
scaledImageRect.origin.x = (newSize.width - scaledImageRect.size.width) / 2.0f;
scaledImageRect.origin.y = (newSize.height - scaledImageRect.size.height) / 2.0f;

UIGraphicsBeginImageContextWithOptions( newSize, NO, 0 );
[image drawInRect:scaledImageRect];
UIImage* scaledImage = UIGraphicsGetImageFromCurrentImageContext();
UIGraphicsEndImageContext();

return scaledImage;

 }

错误信息...

    Error: Bad argument (Wrong input image size. Reason: Training and Test images must be of equal size! Expected an image with 12000000 elements, but got 108000000.) in predict, file /Users/mustafa/Desktop/OpenCVBuild/opencv_contrib/modules/face/src/fisher_faces.cpp, line 140
libc++abi.dylib: terminating with uncaught exception of type cv::Exception: OpenCV(3.4.0-dev) /Users/mustafa/Desktop/OpenCVBuild/opencv_contrib/modules/face/src/fisher_faces.cpp:140: error: (-5) Wrong input image size. Reason: Training and Test images must be of equal size! Expected an image with 12000000 elements, but got 108000000. in function predict

(lldb) 

这给了我一张 9000x12000 的图片（销量增加了 3）。

非常感谢 advice/help。这个问题困扰我好几天了！

Answer 1

我使用本机 OpenCV 调整大小功能解决了这个问题。

 cv::resize(input, output, cv::Size(3000,4000));

输入输出都是cv::Mat

类型