Considering the input vector/matrix D, Matlab commands used to normalize D to zero mean and unit variance are:  

New_Array = (D - mean(D(:))) ./ std(D(:));

Note that once normalized, the distribution of the New_Array would differ from D. You can check it with the command: hist(D); hist(New_Array). Moreover, I read a post in the Matlab forum related with PCA normalization  (attached)

https://es.mathworks.com/matlabcentral/newsreader/view_thread/301665

You can use zscore matlab instruction in order to obtain a shifted and scaled data matrix

PCA does not assume zero mean & unit variance. You can put in all the data you like. However, you can ask yourself what the analysis results reflect? The first PCA factors() will describe those variables that have the largest variance. To avoid this, one can scale the data. In case of your histogram, I assume the units are all the same. So, scaling is not necassary from a certain point of view. However, you could e.g. apply range scaling to increase the weight of the lower values (from your histogram) and decrease the weight of higher values in your PCA analysis. 

you may try our new paper

Article PERCENTILE BASED HISTOGRAM BIN WIDTH