In my experience, the most common approach to find DMR is to use the DSS R package to call and merge DML into regions.
DMC/DML (Differential Methylation Cytosine/Loci) is often calculated by comparing the counts of methylated and unmethylated cytosine at one position between samples.
Steps are to first call DMLs with the callDML() function. Then call DMRs from these first results with callDMR(). Finally annotate the DMLs or DMRs using a GFF file. From there you can look for DMR falling into CpG island. Some of the parameters that can be tuned are the minimum region length, minimum number of differential CpG sites, percentage of CpG sites required to be significant. However I have never heard of what would be the standard values for these parameters. I don't think is it clearly established yet.
So because the process of calling DMLs and DMRs need robust statistical testing, it is suggested to first call the differential methylated states and then annotate it.