How to perform labelling in the 3D extracted features in classification?

I extracted the features from many images in 3D. So the form of the extracted features (for example):

     Image 1 --> 10 x 12 x 30
     Image 2 --> 15 x 10 x 30
     Image 3 --> 13 x 12 x 30
     Image 4 --> 14 x 19 x 30
     .
     .
     So on

In which, the first and second dimension are always changing, but the third dimension is still same (the dimension of the feature).

To do classification, I need to label each extracted features from each image with a label. And to make it easy, I have to keep that features (from each image) in a row.

So I reshape the features into

      feat_reshape= reshape(feature,[size(feature,1)*size(feature,2) size(feature,3)]);

In this case, I'll have 2D features matrix. But still, I need it in 1D.

I already tried using "**imhist**" and keep each feature of each image in each row in a matrix. But still not give a good result. May be because there are many 0 value.

Then I tried to concatenate each column of each features (of each image), so become 1 row. But the problem for this is the different size of column in the concatenate matrix (element in each row).

For example :

      Image 1 --> 10 x 12 x 30 = 3600 elements
      Image 2 --> 15 x 10 x 30 = 4500 elements

Any suggestion,, how I can store my 3D extracted features and then label it to be used in LibSVM MATLAB??